Fortschritte Der Physik - 2015 - Sharpe - Notes On Generalized Global Symmetries in QFT

Fortschritte
der Physik
Fortschr. Phys. 63, No. 11–12, 659–682 (2015) / DOI 10.1002/prop.201500048
Progress
Original Paper
of Physics
Notes on generalized global symmetries in QFT

Eric Sharpe∗
Received 4 September 2015, accepted 4 September 2015

Published online 5 October 2015
There are more 2-groups and higher groups than

It was recently argued that quantum field theories possess one- merely, ‘groups’ of gauge fields and higher-form tensor
form and higher-form symmetries, labelled ‘generalized global potentials (connections on bundles and gerbes), and in
symmetries.’ In this paper, we describe how those higher-form this paper we will give examples of actions of such more
symmetries can be understood mathematically as special cases general higher groups in quantum field theory and string
theory. We will also propose an understanding of cer-
of more general 2-groups and higher groups, and discuss exam-
tain anomalies as transmutations of symmetry groups of
ples of quantum field theories admitting actions of more gen-
classical theories into higher group actions on quantum
eral higher groups than merely one-form and higher-form sym- theories.
metries. We discuss analogues of topological defects for some To be clear, proposals for roles of 2-groups in physics
of these higher symmetry groups, relating some of them to or- have appeared previously in a number of papers, in e.g.
dinary topological defects. We also discuss topological defects [2–5], in discussions of the String group (see e.g. [6–9]),
in cases in which the moduli ‘space’ (technically, a stack) ad- in the Yetter model (see e.g. [10, 11]), and in lattice gauge
mits an action of a higher symmetry group. Finally, we outline theories (see e.g. [12, 13]), to name a few examples. Ad-
a proposal for how certain anomalies might potentially be un- ditional categorical generalizations of (orbifold) groups,
via an application of defects to generalize the ordinary
derstood as describing a transmutation of an ordinary group
orbifold construction, are discussed in [14–19]. The pur-
symmetry of the classical theory into a 2-group or higher group
pose of this paper is to merely to link the recent work [1]
symmetry of the quantum theory, which we link to WZW mod- to other work on 2-groups, to review a few highlights, and
els and bosonization. to provide a few hopefully new results, proposals, and
applications.
We begin in section 2 with a brief introduction to
2-groups and higher groups, omitting mathematical de-
1 Introduction tails for the sake of readability. In section 3 we describe
several examples of quantum field theories admitting
The recent paper [1] proposed a more general class of higher group symmetries. We begin with a brief review
symmetries that should be studied in quantum field of the gauge theory phase analysis of [1], then describe
theories: in addition to actions of ordinary groups, symmetries of gauge theories with massless matter that
it proposed that we should also consider ‘groups’ of is invariant under a finite subgroup of the gauge group.
gauge fields and higher-form analogues. For exam- (Such theories have been studied in a number of papers,
ple, Wilson lines can act as charged objects under and in two dimensions naturally decompose into dis-
such symmetries. By using various defects, the paper joint unions of theories, as is reviewed in that section.)
[1] described new characterizations of gauge theory We also review how boundary structures in Dijkgraaf-
phases. Witten theory and also WZW models define further ex-
Now, one can ask to what extent it is natural for n- amples of theories admitting higher group symmetries,
forms as above to form a group. In particular, because albeit more complicated higher groups than just p-form
of gauge symmetries, the group multiplication will not symmetries. Both those examples of higher group sym-
be associative in general, unless perhaps one restricts to metries are discussed elsewhere in the literature, but
suitable equivalence classes, which does not seem nat-
ural in general. A more natural understanding of such
symmetries is in terms of weaker structures known as Department of Physics MC 0435, 850 West Campus Drive, Virginia
2-groups and higher groups, in which associativity is Tech, Blacksburg, VA 24061
weakened to hold only up to isomorphisms. ∗ Corresponding author E-mail: ersharpe@vt.edu

C 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Wiley Online Library 659
Fortschritte
der Physik
15213978, 2015, 11-12, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/prop.201500048 by National University of Science and Technology, Wiley Online Library on [09/07/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
E. Sharpe: Notes on generalized global symmetries in QFT
Progress
Original Paper
of Physics
are reviewed here as they play a role later. At the end and identity relations
of that section, we also speculate on how current al-
gebras in higher dimensions may define further exam- x · x−1 = 1, x−1 · x = 1
ples of higher group symmetries, and also mention some
other potential applications. In section 4 we briefly dis- are also replaced by isomorphisms. These isomorphisms
cuss cosmological defects in the context of higher group must satisfy relations of the form (see e.g. [23])
symmetries, as well as ‘generalized moduli spaces’ (tech-
nically, stacks), defined here as ‘spaces’ which admit ac-
tions of higher groups. Finally, in section 5 we conjecture
that some anomalies might be interpreted as promoting
classical ordinary-group symmetries to quantum higher-
group symmetries.
One of the ideas that drives this paper is that when and
the action of an ordinary group G is broken – in mathe-
matics, by trying to lift G to a bundle that does not ad-
mit a G action, say, or in physics, via anomalies – there
is often a larger or higher group that will act instead. We
will encounter various manifestations of this principle in
There are technical distinctions between various de-
several contexts.
scriptions of 2-groups, which will not be relevant for our
An idea that has often been repeated is that one
purposes here (see instead e.g. [23]).
should look for fundamental symmetries to better un-
The description above suggests that it may be possi-
derstand string theory (see for example [20] for one well-
ble to construct 2-groups from a given group by weak-
known example). The philosophy espoused by this paper
ening associativity by an automorphism defined by an
and others mentioned above is that perhaps instead of
element of degree 3 group cohomology, and in fact,
only looking for ordinary group symmetries, we should
this is the case, see e.g. [23][section 8.3], [25, 26]: for a
also look for higher group symmetries.
given group G, group cohomology H 3 (G, U(1)) (with triv-
ial action on the coefficients) defines a corresponding 2-
group. Briefly, the basic idea is that a 3-cocycle α(g, h, k)
2 Overview of 2-groups and higher groups defines an isomorphism
In this section, to make this paper self-contained, we ∼

(g · h) · k −→ g · (h · k)
will outline definitions and examples of 2-groups and
higher groups, as these ideas are not widely known in and as the cocycle is closed in group cohomology, the
the physics community. We will begin with 2-groups (see isomorphisms so defined satisfy the pentagon identity:
e.g. [6, 8, 21–25] for discussions and introductions to 2-
groups). α(g, h, k) α(g, hk, ) α(h, k, ) = α(gh, k, ) α(g, h, k).
Briefly, given objects x, y, z in a 2-group, instead of
demanding that multiplication be strictly associative We shall return to this example shortly.
Let us next discuss 2-groups appearing in discussion
(x · y) · z = x · (y · z), of “1-form symmetries.” For G abelian, we can define [27]
the 2-groups BG, BGconn , and B Gconn , which we will de-
we instead demand that the two sides be merely isomor- fine over the next several paragraphs. The first 2-group
phic, related by isomorphisms BG assigns1 to any manifold the category of principal G
∼ bundles. There exists a tensor product that tensors any
α(x, y, z) : (x · y) · z −→ x · (y · z).
Similarly, the identity axioms

1 Technically, in our conventions we will take BG to be the stack
1 · x = x, x · 1 = x [point/G]. More generally, classifying stacks and classifying
spaces naturally admit the structure of higher groups. We will
are also replaced by isomorphisms largely avoid using the language of stacks in this paper so as to
make it more accessible, aside from technical footnotes of this
∼ ∼
x : 1 · x −→ x, rx : x · 1 −→ x form.
660 www.fp-journal.org
C 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Fortschritte
der Physik
Fortschr. Phys. 63, No. 11–12 (2015)
Progress
Original Paper
of Physics
two G bundles to produce another G bundle2 , and the Finally, let us discuss B Gconn . This is the analogue of
resulting operation defines a multiplication that is asso- BGconn for flat bundles. In other words, B Gconn assigns
ciative up to higher homotopies, as above. the category of flat G bundles with connection. If G is fi-
The 2-group BGconn is the analogous classifying ob- nite, then for trivial reasons
ject for principal G bundles with connection, which ten-
sors bundles and adds the connections. (If G is finite, B Gconn = BG = BGconn .
then as there is no nontrivial connection on a princi-
pal G-bundle, BGconn = BG.) This 2-group appeared im- Next let us turn to further examples of 2-groups,
plicitly in [1]. Since the isomorphisms must preserve which cannot be understood as simply as “1-form sym-
connections and so are more restricted than the isomor- metries.” In particular, we will next consider examples of
phisms in BG above, one might worry that the result- 2-groups formed as extensions, as described in e.g. [6].
ing 2-groups would all be equivalent to ordinary groups. For G a group, there are several 2-group extensions G̃
However, this is not the case. For example, discrete tor- given by3
sion implicitly provides examples of 2-groups of this
form which are not equivalent to ordinary groups. As dis- 1 −→ BU(1) −→ G̃1 −→ G −→ 1,
cussed in e.g. [29], discrete torsion arises from choices 1 −→ BU(1)conn −→ G̃c −→ G −→ 1,
of equivariant structures on B fields – one combines the
action of the orbifold group on the underlying space 1 −→ BU(1) −→ G̃0 −→ G −→ 1,
(here, a point) with gauge transformations of the B field.
differing by whether one extends by a stack BU(1),
The gauge transformations, themselves principal U(1)
BU(1)conn , or a classifying space BU(1). (One can also
bundles with connection, are required to obey the group
consider extensions by the various B U(1)conn ’s, but the
law up to isomorphism – in essence, a 2-group struc-
examples above will suffice for this paper.) Extensions of
ture – and those isomorphisms are encoded in elements
the form G̃1 are discussed in [6]. In particular, this exten-
of H 2 (, U(1)), as connection-preserving gauge transfor-
sion can be understood as a U(1)-gerbe over G, with suit-
mations of a principal U(1) bundle.
able multiplicative structure. For G finite, these are pre-
As 2-groups, BG and BGconn are not equivalent to one
cisely the 2-gerbes discussed above which are classified
another [26, 27]. Now, that said, to classify 2-groups topo-
by group cohomology H 3 (G, U(1)).
logically, we will sometimes replace 2-groups by their
Let us now suppose instead that G is simple and
‘geometric realization.’ Ultimately because the space of
simply-connected; in this case, extensions of this form
connections on a fixed bundle is contractible, the geo-
above are classified by H 3 (G, Z) (see appendix A). (Each
metric realization of BG is homotopic to that of BGconn ,
type of extension is classified in the same way, as dis-
despite the fact that the 2-groups are not equivalent [30].
cussed in appendix A.) Now, for G simple and simply-
In both cases, the geometric realization is homotopic to
connected, H 3 (G, Z) = Z, and so possible extensions are
the ordinary classifying space for G, which we will denote
classified by an integer. The reader may find this struc-
as BG (with a non-boldface B), to distinguish it from the
ture reminiscent of WZW models, and indeed there is a
2-groups above. (See for example [31] for a basic discus-
close connection: the underlying U(1)-gerbe appearing
sion of the classifying space, and appendix A for refer-
in WZW models, for which the Wess-Zumino term acts as
ences on homotopies of geometric realizations.)
a coupling to the B field, is precisely the 2-group G̃1 above
Now, if one lifts to a loop space, some of this structure
(with suitable multiplication), and the integer classifying
simplifies. For example, G-gerbes over a space X become
the extension is the same as the level of the WZW model.
principal G-bundles on the loop space LX . In particular,
We will return to this connection later.
a Wilson line is effectively a function on LX , and so obeys
So far this discussion has been rather abstract. Let us
an ordinary group law. Thus, if one is working with these
now try to make it more concrete by discussing more ex-
structures exclusively through Wilson lines, then at least
plicitly the multiplication on a 2-group G̃, built as an ex-
to some extent it is entirely reasonable to speak about
tension of (simple, simply-connected) G by some version
them as ordinary groups (involving Wilson lines) rather
of BU(1). This product is more easily described on the
than higher groups, as in fact was done in [1].
loop space of G, or rather a minor variation of the loop
2 Technically, we should describe this in terms of tensor prod-

ucts of torsors [28][section 5.1], [2], but for the purposes of this 3 Further discussion of the relation between these extensions
paper, we shall be content with looser language. can be found in e.g. [32].

Fortschritte
der Physik
Progress
Original Paper
of Physics
space. (The resulting description will also implicitly en- of Q by

1 :
code a version of Wilson lines for 2-groups defined as ex-
tensions G̃ of the form above.) 0 −→
1 −→ Q −→ Q −→ 0
First, let us describe elements of G̃. Following
[28][section 6.4], we can describe them as certain equiv- is given by the first Pontryagin class p1 (P) [33]. The rel-
alence classes of pairs (σ, z), where σ : D2 → G for D2 a evance of this construction to the Green-Schwarz con-
two-dimensional disk (corresponding to a filled-in loop) dition and associated anomalies in heterotic strings has
and z ∈ U(1) (corresponding to an ordinary U(1) Wilson been discussed in e.g. [34–36]; see also [37] for a discus-
line). Then the product of two elements of G̃ (over this sion of the Green-Schwarz anomaly in present language.
analogue of the loop space) is defined by In principal, n-groups for n > 2 can be defined sim-
ilarly to 2-groups, by breaking the relations above to
only hold up to higher levels of isomorphisms, which
∗
(σ1 , z1) · (σ2 , z2) = σ1 σ2 , z1 z2 exp −2πi (σ1 × σ2 ) ω , themselves obey higher identities. For example, to get a
D2
3-group, one would replace the identity
where ω is a 2-form on G × G, defined as follows. If ν is
α(w, x, y · z) ◦ α(w · x, y, z)
the a multiple of the canonical 3-form on G (associated
with the Wess-Zumino term in WZW models, and encod- = (1 ⊗ α(x, y, z)) ◦ α(w, x · y, z) ◦ (α(w, x, y) ⊗ 1)
ing the level), p1,2 : G × G → G are projection maps, and
m : G × G → G the ordinary multiplication in G, then with isomorphisms
β(w, x, y, z) : α(w, x, y · z) ◦ α(w · x, y, z)

p∗1 ν + p∗2 ν − m∗ ν = dω.
∼
−→ (1 ⊗ α(x, y, z)) ◦ α(w, x · y, z) ◦ (α(w, x, y) ⊗ 1)
(Intuitively, we can think of σ as a semiclassical state of
the string in a WZW model, and the phase acquired in which themselves obey higher identities.
multiplication as the integral of the Wess-Zumino term The theory of n-groups seems to be less well de-
over the path swept out by the change in region.) As this veloped than that of 2-groups; we refer the reader to
is defined over (a version of) the loop space, rather than e.g. [38–40] for a few details. Briefly, the complications
over G itself, this is an honest group structure: associativ- in making sense of n-groups are tied into the com-
ity holds on the nose, rather than up to cocycles. In more plications of understanding higher categories, and in
detail, associativity of the multiplication follows from the modern language, are perhaps best understood by
fact that for any σ1,2,3 , working with ∞-categories (and hence ∞-groups).
We will not need that level of technology, but we will

occasionally make conjectures based formally on
(σ2 × σ3 )∗ ω + (σ1 × σ2 σ3 )∗ ω
n-groups, so we will describe a few examples we shall
use in this paper, and for technical definitions we refer
= (σ1 × σ2 )∗ ω + (σ1 σ2 × σ3 )∗ ω, interested readers to e.g. [40].
In that spirit, for G abelian, we will define [27] Bq G
as follows from the fact that these four terms correspond to associate to any manifold the G (q − 1)-gerbes on that
schematically to the four faces of a tetrahedron, as illus- manifold4 , Bq Gconn the G (q − 1)-gerbes with connec-
q
trated schematically below: tion, and B Gconn the flat G (q − 1)-gerbes with connec-
tion. If G is finite, then for trivial reasons
q
B Gconn = Bq G = Bq Gconn .
To make our notation uniform, we shall define B0 G to

be the ordinary group of smooth maps from the mani-
fold into G, and B0 Gconn the ordinary group of constant
maps into G. (The classifying space for the former is G;
for the latter, K (G, 0).) We then recognize that in more
In passing, we should mention that there is a related
story for Lie algebroids. Given a Lie algebroid Q associ-
ated to some bundle P, the obstruction to lifting to a non- 4 Readers familiar with the bar construction of classifying spaces
trivial Courant algebroid Q determined by an extension will recognize the deliberate notational parallel.
Fortschritte
der Physik
Fortschr. Phys. 63, No. 11–12 (2015)
Progress
Original Paper
of Physics
typical symmetry discussions in QFT, B0 Gconn defines a B U(1)conn around the path:
global symmetry group, and B0 G a local symmetry group.
Analogous parallels hold for q > 0. exp A · ds →
exp A · ds exp A0 · ds .
Analogues of the extension construction above exist
for higher groups [27, 40, 41]. For example, one can con-
Mathematically, this can be understood in terms of pass-
sider extensions of the form
ing to the (based) loop space: line bundles on a given
space become functions on loop space, and B U(1)conn
1 −→ Bk U(1) −→ G̃1,k −→ G −→ 1,
becomes B0 U(1)conn , the group of constant U(1) ele-
1 −→ Bk U(1)conn −→ G̃c,k −→ G −→ 1, ments, on the loop space, so that overall this one-form
symmetry reduces to an ordinary (zero-form) group
1 −→ Bk U(1) −→ G̃0,k −→ G −→ 1,
symmetry on the loop space. (That said, Wilson lines
themselves can at least sometimes define a 2-group, see
differing by whether one extends by a higher stack
e.g. [46][section 3].)
Bk U(1), Bk U(1)conn or a space Bk U(1). As outlined in
If a Wilson line has a nonzero vev in the limit of large7
appendix A, these are classified by H k+2 (G, Z) for Lie
loop sizes, then the one-form B U(1)conn symmetry said
groups G of nonzero dimension, and group cohomology
to be spontaneously broken. Conversely, if one knows
H k+2 (G, U(1)) for finite G.
that a symmetry of this form is unbroken, one can argue
that Wilson loop vevs in large loop limits must vanish, a
selection rule on Wilson lines discussed in [1].
3 Examples of higher group symmetries in QFT
Reference [1] observes that the gauge field Aμ is natu-
rally understood as the Goldstone boson associated with
In this section we list some examples of higher group
that spontaneously broken symmetry, as it undergoes a
symmetries in quantum field theories.
translation under the symmetry group action.
To close the intellectual loop, it would be helpful to
see explicitly that the state space of an abelian gauge the-
3.1 Review of q-form symmetries in gauge theories
ory has the structure of a 2-vector-space in a represen-
tation of B U(1)conn . At a very formal level, such a de-
To make this paper self-contained, let us begin with a
scription was proposed in [47], which we shall outline
very brief outline of the highlights of some of the exam-
here8 . (Readers unfamiliar with stacks may wish to skip
ples of global q-form symmetries (special cases of higher
the rest of this paragraph.) Briefly, the idea is to think of
group symmetries) discussed in [1].
states in abelian gauge theory on a manifold X (expand-
Reference [1] gives a number of examples5 of global
ing about an arbitrary principal U(1) bundle with con-
q-form symmetries involving shifts of an existing gauge
nection) as defined by wavefunctions, defined on a sub-
field. The prototype for many of their examples was
set of [X, BU(1)conn ] satisfying a polarization condition.
a four-dimensional U(1) gauge theory on some gen-
Using “crit” to label the pertinent subset, the wavefunc-
eral four-manifold. This admits an (electric) action6 of
tions are elements of the 2-vector-space
B U(1)conn defined by shifting the gauge field A by a
flat U(1) connection A0 , as A
→ A + A0 , and tensoring
[[X, BU(1)conn ]crit , C] ,
the gauge bundle by the corresponding flat U(1) bun-
dle. There is also a dual magnetic action of B U(1)conn , which naturally admits an action of B U(1)conn [27].
defined by shifting the magnetic potential in the same More globally, we are under the general impression that
fashion. the representation theory of higher groups is still un-
Note that under the action above, Wilson lines are der development, so in the rest of this paper we will
charged, picking up the holonomy of the element of
7 The reader might well observe that any nonzero vev of an oper-
5 See also [42–45] for related ideas. ator that transforms under a symmetry defines a spontaneous
6 In more formal language [27], we could describe the gauge field symmetry breaking. Here, however, we intend to refer to the
A and corresponding bundle via a map into the classifying phase visible in the IR, for which the limit of large loop size is
stack BU(1)conn . Now, BU(1)conn admits an action of itself, relevant. In any event, we follow the convention set in [1].
and in particular the substack B U(1)conn , which is the action 8 We thank U. Schreiber for outlining his argument to us, which
being described above. we repeat here.

Fortschritte
der Physik
Progress
Original Paper
of Physics
occasionally make physics-based conjectures regarding theory; as one increases its size, one can excite
representations, but for the most part we will not be arbitarily massive matter, and so the theta angle peri-
able to make strong statements of results concerning odicity provides a test for charged matter beyond the
representations. renormalization cutoff scale. (For more information,
Some related examples discussed in [1] include: see for example [50–52], where this model was origi-
r A four-dimensional G gauge theory admits a global r
nally discussed.)
In principle, one could use defects to distinguish
action of B Z(G)conn defined by shifting the gauge
these theories. For example, one could specify that
field by a flat9 Z(G) connection, and similarly tensor-
the theory contains Wilson lines in the charge 1 rep-
ing the gauge bundle by the corresponding flat Z(G)-
resentation. Such Wilson lines would not be well-
bundle.
r A theory with a dynamical q-form potential admits
defined after dividing the charges by k, and so speci-
q fying their existence would distinguish these theories.
a global action of B U(1)conn given by shifting the
Such structures were discussed in four-dimensional
q-form potential by a flat q-form, and tensoring the
theories in the context of discrete theta angles in [53,
q-gerbe by the corresponding flat q-gerbe. (See also
54].
[48] and references therein for related examples.) r In principle, the global structure of the gauge group
In all these cases, various defects10 play a role analogous could also be detected via gauge transformations on
to charged particles for ordinary symmetries. non-simply-connected spacetimes. Put another way,
In any event, the paper [1] focused on analyzing if the spacetime is topologically nontrivial, then to
gauge theory phases using the ideas above, whereas in specify matter fields, one must specify a bundle to
this paper we will look in different directions. which the matter couples, not just a representation.
The specification of the bundle eliminates any am-
biguities, and as different bundles lead to different
3.2 Gauge theories with subgroup-invariant massless zero modes and different anomalies, one also sees
matter distinct physics. This is also discussed in for example
[50–52].
In this section we shall study gauge theories in which
a finite subgroup of the center of the gauge group acts At low energies, below the scale of any massive min-
trivially on massless matter, and related orbifolds, as fur- imally charged matter, this theory has a BZk symme-
ther examples of theories with generalized global sym- try, acting by translations of the U(1) gauge field in a Zk
metries. subgroup, as this leaves the action invariant. (This same
As a warm-up, consider the two-dimensional su- group may act nontrivially on defects that distinguish the
persymmetric CPn model, consisting of a U(1) gauge charge k and charge 1 massless matter cases; however,
theory with n + 1 charged massless chiral multiplets, a nontrivial action on such operator vevs should be in-
but rather than giving all the chiral multiplets charge terpreted in terms of spontaneous symmetry breaking,
1 as in the usual construction, instead give them all rather than explicit symmetry breaking.) This BZk action
charge k. Perturbatively, this would appear to be the is of the same form as discussed in the last section: the
same as the ordinary CPn model, in which all mat- gauge bundle is ‘tensored’ with a Zk bundle, for example.
ter fields have charge 1, but in fact these cases can be However, at higher energies, this BZk symmetry is explic-
distinguished: itly broken by any massive minimally charged matter, as
r One way to distinguish these theories is by adding
the action is no longer invariant.
We could also discuss the theory without massive
massive fields of charge ±1 to the theory with mass-
minimally charged matter or defects fixing the global
less charge k matter. The existence of this massive
structure of the gauge group. This theory, in which the
matter can still be detected even below the cutoff
charges of all matter fields are multiples of k, naively ap-
scale, via the periodicity of the theta angle, which
pears to also have a BZk symmetry, but this symmetry is
acts as an electric field in two dimensions. To see
an artifact of the charge scaling, and so has no physical
this, simply build a capacitor in the two-dimensional
significance.
Another set of examples involves orbifolds by finite
9 Z(G) denotes the center of G. groups that do not act effectively on the target space.
10 See for example [49] for a discussion of defects in topological A simple example, discussed extensively in e.g. [51], is
field theories. the orbifold [X/D4 ], where D4 is the eight-element group
Fortschritte
der Physik
Fortschr. Phys. 63, No. 11–12 (2015)
Progress
Original Paper
of Physics
described as an extension as
1 −→ Z2 −→ D4 −→ Z2 × Z2 −→ 0, (g, h)
→ (gzm , hzn ).
and where D4 acts on X by first projecting to Z2 × Z2 , Since the Z2 acts trivially on X ,

so that the remaining Z2 acts trivially on X . By com-
puting partition functions, it is simple to demonstrate Zgzm ,hzn = Zg,h ,
that this theory is not the same as the [X/Z2 × Z2 ] orb-
ifold – both theories admit modular-invariant partition and so the partition function is preserved.
functions, but those modular-invariant partition func-
tions are different. In particular, if H is a subgroup of More generally, for any G orbifold of a space X , if a sub-
the orbifold group that acts completely trivially on the group H acts trivially on X , then in the same fashion as
space, then the theory admits a BH action. In this case, above, BH defines a symmetry of the theory, as can be
the [X/D4 ] orbifold admits a BZ2 action. seen in rotations of the G bundles by H subbundles.
It is straightforward to see the BZk actions in orbifold Four-dimensional analogues were discussed in [55].
examples, but for completeness, we shall work through Two examples of this form are as follows:
two examples here. r A U(1) gauge theory with N matter fields of charge +k
r First, consider an orbifold [X/Zk ] where all of the Zk and N matter fields of charge −k.
acts trivially on X . In other words, for all x ∈ X and r An SU(2) gauge theory with adjoints.
all g ∈ Zk , g · x = x. The one-loop partition function
of this theory is involves a sum over principal Zk bun- At low energies, generically along the Higgs branch, the
dles, of the form second reduces to a U(1) gauge theory with matter of
charge divisible by 2, i.e. an example of the former the-
1 ory. As in two dimensions, these four-dimensional theo-
Z= Z(X ) = |Zk |Z(X ).
|Zk | ries can be distinguished from minimally-charged mat-
g,h∈Zk
ter theories in several ways:
(A principal Zk bundle on T 2 is specified by a com- r One option is to add massive minimally-charged mat-
muting pair of elements of Zk , hence the sum over
ter to the theory, as in two dimensions. The presence
g, h ∈ Zk .) The action of BZk is merely to rotate the
of such matter can no longer be sensed by the theta
bundles amongst themselves. For example, given an-
angle periodicity, as the theta angle no longer acts as
other bundle defined by a (commuting) pair (g , h ),
an electric field, but instead in a theory coupled to
the action on the bundle defined by the pair (g, h) is
gravity, one can use Reissner-Nordstrom black holes
merely
to similar effect, as discussed in [55].
r Another option is to specify a set of defects in the
(g, h)
→ (g g, h h).
theory which are well-defined only for certain global
r A somewhat less trivial example is provided by the gauge groups. This was the strategy followed in [53,
[X/D4 ] example given above. This theory admits a BZ2 54] to distinguish SU(2) from SO(3) theories, for
action, which we can see in the one-loop partition example.
function as follows. Write the one-loop partition func- r Finally, as before, if the spacetime is topologically
tion as nontrivial, then a unique specification of the matter
will have the same ffect.
1
Z= Zg,h (X ),
|D4 | In two dimensions, gauge theories in which a sub-
g,h∈D4 ,gh=hg
group of the gauge group acts trivially on massless mat-
where the sum is over commuting pairs of elements ter are equivalent to theories with restrictions on non-
of D4 (i.e. principal D4 bundles on T 2 ), and Zg,h (X ) perturbative sectors. In four dimensions, the analogue is
denotes the partition function with boundary con- more complicated. This matter is discussed in greater de-
ditions determined by g, h. Now, a Z2 bundle is detail in [55, 57]. These theories can also be understood,
termined by a pair of elements of Z2 ⊂ D4 , whose in different language still, in terms of QFT’s coupled to
elements we will denote {1, z}, where z2 = 1. Thus, a TFT’s, though we shall not use that language here.
general Z2 bundle is given by a pair (zm , zn ) for inte- There are many other related examples, in which
ger m, n. Such a bundle will act on a D4 bundle (g, h) massless matter is invariant under a subgroup G of the

Fortschritte
der Physik
Progress
Original Paper
of Physics
center of the gauge group, and these examples all have a references and reviews). It would be interesting to un-
BG symmetry at low energies. derstand if there are analogues of decomposition for any
These theories are sometimes known as gerbe theo- notion of BG gauge theories in some dimension. In any
ries, because they are typically sigma models on gerbes event, reference [55] contains a study of how these gerbe
as discussed in, for example, [52, 56]. Very briefly, for structures over moduli spaces vary under dualities.
those readers who are curious, a sigma model on a
(Deligne-Mumford) stack is defined by first picking a pre- 3.3 Boundary structures in Dijkgraaf-Witten theory
sentation of the stack as a global quotient [X/G] of some
ordinary space X by some group G, which need not be Recall that for finite G, an element of group cohomol-
finite and need not act effectively (but whose stabilizers ogy H 3 (G, U(1)) (with trivial action on the coefficients)
on X are finite). The ‘sigma model on the stack’ is then defines a 2-group G̃ as the extension
a G-gauged sigma model on X , or rather, its universality
class. If a subgroup H of G acts trivially on X , then the 1 −→ BU(1) −→ G̃ −→ G −→ 1.
stack is known technically as an H gerbe, and the theory
is a sigma model on a gerbe. In this section, we shall review arguments that this 2-
group G̃ acts on the boundaries in three-dimensional
We have already seen how theories of this form ad-
Dijkgraaf-Witten theories [69] defined by the corre-
mit BH actions – by rotating principal G bundles by H
sponding element of H 3 (G, U(1)). (We emphasize that
subbundles, which (as H acts trivially) leaves the the-
the observations of this subsection are not original to us,
ory invariant. (This is true for both gauge theories and
but instead are reviewed for completeness and later use
orbifolds, and we have seen examples of each.) As sigma
in this paper. See instead e.g. [40][p. 757] for references.)
models on gerbes, we can view these BH structures an-
First, consider the special case of a trivial G gauge
other way. An H-gerbe (for H finite) over a space is, math-
theory in two dimensions, for G finite – a G-orbifold –
ematically, the total space of a BH bundle. As such, BH
with discrete torsion. Discrete torsion is defined by an
acts on the fibers of the bundle, and so acts on the gerbe
element of H 2 (G, U(1)) (with trivial action on the coef-
and, physically, on the sigma model.
ficients), and elements of that same group cohomology
The same language gives another perspective on
group determine a central extension G̃:
sigma models on gerbes. As has been discussed else-
where (see e.g. [55]), in two dimensions these are 1 −→ U(1) −→ G̃ −→ G −→ 1.
equivalent to sigma models with restrictions on non-
perturbative sectors. One way to think about the origin As was argued in [29, 70–72], the extension G̃ acts on the
of those restrictions is as a requirement that the sigma Chan-Paton factors and D-branes. The extension G̃ is still
model maps preserve the BH invariances. an ordinary group, but this example provides a prototype
The Higgs moduli spaces of these theories also have for the 2-group action that will appear on boundaries of
gerbe structures, and so admit actions of higher groups. three-dimensional Dijkgraaf-Witten theories.
We will explore this point in greater detail in section 4.2, Abstractly, to define the two-dimensional G gauge
where such moduli spaces will be described as ‘general- theory in the presence of the boundary, one must ex-
ized moduli spaces.’ press how the group G acts on the Chan-Paton factors
So far we have discussed some basic examples. It has as encoded in a Wilson line along the boundary (or
been shown that dualities break the BG symmetries dis- equivalently a bundle L with connection). If the G ac-
cussed in this section. Specifically, it has now been estab- tion lifts honestly, meaning that there are isomorphisms
lished that these two-dimensional theories ‘decompose’ ψg : g ∗ L → L such that the ψg ’s obey the group law,
(via a form of T-duality) into disjoint unions of simpler then L is said to admit a G-equivariant structure.
theories, without these BG symmetries, which solves a In general, however, the group law might not be
technical issue with cluster decomposition, see e.g. [56] obeyed, meaning the lift of G to the boundary the-
for a discussion of decomposition in two-dimensional ory is obstructed. Such obstructions are encoded in el-
nonlinear sigma models, [57] for a discussion of de- ements11 of H 2 (G, U(1)). In such a case, although the
composition in two-dimensional nonabelian gauge the- action of G on the boundary theory is obstructed,
ories, and [58] for a heterotic version. This decompo- there is an action of G̃ (the lift associated with the
sition makes predictions for Gromov-Witten invariants,
which have been checked rigorously (see e.g. [59–64]),
and also plays a role in understanding phases of certain 11 In this context, H 2 (G, U(1)) is known as Mumford’s group of
GLSMs [56, 65–67] (see [68] for a more complete list of L or the theta group of L [73].
Fortschritte
der Physik
Fortschr. Phys. 63, No. 11–12 (2015)
Progress
Original Paper
of Physics
boundaries is given in [69][equ’n (6.35)] as
α(g, h, k)α(h, k, g)α(k, g, h)

.
α(g, k, h)α(h, g, k)α(k, h, g)
This is the same phase factor that was computed for for-
mal C-form analogues of discrete torsion in [74]. It is in-
Figure 1 A 3-simplex. variant under group coboundaries, as well as SL(3, Z).
The interpretation of the 3-cocycle α(g, h, k) can be read
off from12 [74]: it is precisely the obstruction to associa-
same element of H 2 (G, U(1))) on the boundary the- tivity in the BU(1)conn transformations. Finally, note that
ory: the boundary theory is always G̃-equivariant, although the nonlinear sigma model M played a role in
for the extension determined by the obstruction to initially setting up our discussion, it is irrelevant to the
G-equivariance. result.
There is a closely analogous story in three- Putting this together, we see that in Dijkgraaf-Witten
dimensional Dijkgraaf-Witten theories [69]. To help theory on a manifold with boundary, the gauge group G
clarify the discussion, we shall assume that the three- can not act honestly on the boundary theory, but rather
dimensional finite G gauge theory includes a trivial associativity of its group law is obstructed by an ele-
action on a nonlinear sigma model on some space M. (It ment of group cohomology H 3 (G, U(1)), the same ele-
will become clear that M is irrelevant.) Then, a boundary ment defining the Dijkgraaf-Witten action.
in the theory should have an action which includes a Now, although G itself can not act honestly on the
term of the form boundary theory, the element of H 3 (G, U(1)) defines a 2-
group extension
φ∗B
1 −→ BU(1) −→ G̃ −→ G −→ 1,
for some two-form potential B, and map φ from the which does act on the boundary theory – the obstruction
three-dimensional spacetime into M. Assuming that the to associativity indicates that one should replace G by G̃
boundary of the boundary is empty, this theory has a nat- above.
ural action of the 2-group BU(1)conn , which acts on B by It is worth mentioning that a closely related argument
gauge transformations of the familiar form B
→ B + d. for ordinary discrete torsion in two dimensions appears
However, this is only part of the 2-group that acts on the in [72]. There, the fact that the group 2-cocycle defines a
boundary of this theory. boundary phase factor was applied to derive an expres-
Since the bulk contains a G gauge theory, to fully sion for a 2-simplex
describe the theory, we must describe how G acts on
the boundary. In general, G can be combined with
gauge transformations of the B field, i.e. elements of
BU(1)conn . As discussed elsewhere, however, associativ-
ity in BU(1)conn holds only up to isomorphism. If the iso-
morphisms cannot be trivialized, then we do not have which was then applied to compute phase factors for
any sort of homomorphism G → BU(1)conn , but rather a Riemann surfaces of various genera, obtaining results
more complicated structure comes into play. matching standard discrete torsion phase factors. The
We can link the breakdown in associativity in those conclusion there was analogous: the same element of
boundary gauge transformations to the group cohomol- H 2 (G, U(1)) that defines discrete torsion, also defines
ogy defining the bulk action. Recall from [69][section 6.5] an extension of G that acts honestly on the boundary
that the action associated to a 3-simplex such as that theory. In that case, the extension is an ordinary group
in figure 1 is given by a group cohomology 3-cocycle
α(g, h, k). In fact, by imagining that the 3-simplex con-
tains a boundary just as in [72], we can identify the 12 Although [74] was written to describe group actions on C fields,
boundary phase with α(g, h, k). with only minor modifications it applies to the present case,
The total action in the case that spacetime is a three- essentially by identifying the B fields for various g as the im-
torus, for example, can be computed from a simplicial age of the boundaries, rather than as gauge transformations
decomposition of T 3 , and for fixed group elements on the themselves.

Fortschritte
der Physik
Progress
Original Paper
of Physics
extension, whereas here the extension that acts on the where the choice of extension of G by BU(1) is the level
boundary theory, is a 2-group extension. of the WZW model. In this case, the stack underlying the
The same argument appears to imply that boundaries 2-group is the U(1) gerbe whose connection B is defined
of higher-dimensional analogues of Dijkgraaf-Witten above.
theory, defined by elements of H d (G, U(1)) in d bulk di- As a quick consistency check, note that such exten-
mensions, should be acted upon by (d − 1)-groups given sions G̃1 are classified by (appendix A) elements of H 3 (G),
as extensions which also classify the allowed levels of the WZW model
(for simple simply-connected G, which we assume).
1 −→ Bd−2 U(1) −→ G̃ −→ G −→ 1, This 2-group is merely acting by automorphisms
of the underlying gerbe. In this case, the 2-group is
Indeed, precisely this structure has been discussed else-
not merely “G-valued one-forms,” but rather is more
where, see for example [40][section 3.9.130].
complicated.
A different approach to Dijkgraaf-Witten theory and
Physically15 , we can describe the action of G̃1 by right-
2-groups is discussed in [11]. The approach there seems
or left- multiplication on the fields of the WZW model as
to be to extend Dijkgraaf-Witten theory to 2-groups,
follows. First, describe a multiplication on G̃1 , i.e. a map
rather than consider boundaries in Dijkgraaf-Witten the-
G̃1 × G̃1 → G̃1 , by a pair (h, A) consisting of a group ele-
ory, as we have reviewed here. A related recent paper [75]
ment h ∈ G and a connection A on a principal U(1) bun-
describes Bp G gauge theories, for G finite.
dle over G. (The effect of multiplication by an element of
G̃1 is to induce an automorphism of G̃1 , and such auto-
3.4 WZW models morphisms are essentially specified by pairs of the given
form.) To describe the action on the WZW model, in
As seems to be very well-known in certain circles (see
closed string path integral quantization, we must specify
e.g. [8, 40]), a different example of a theory with a type
the action on the classical map g : → G ( the world-
of one-form (2-group) action is provided by WZW mod-
sheet) and on the WZW model B field. These actions are
els [76]. For applications later in this paper, we shall give
as follows:
a physics-oriented review of the highlights here. Recall
that a WZW model in two dimensions is essentially a r g
→ hg or gh (depending upon whether a left- or
sigma model on a group manifold G with a B field with right-multiplication was specified),
nonzero curvature, where the B field is encoded in the r B
→ B + F, where F is the curvature of A.
Wess-Zumino term via its curvature as
The action on g itself is identical to the G action in
g ∗B = Tr (g −1 dg)3 an ordinary WZW model, and has been extensively dis-
Y cussed elsewhere. The action on B leaves the curvature
for a three-manifold Y bounding . (Strictly speaking, of B invariant, and hence the Wess-Zumino term is un-
the Wess-Zumino term is defined up to addition of a affected. In principle, the B field periods have changed

closed13 2-form; locally, by g ∗ F, but since F lies in integral cohomology, and the
B field curvature is multiplied by an integer (the level),
∗
g B∼ g ∗ (B + d) the closed string theory is invariant.
Put more simply, the action of G̃1 on the fields of the
for some one-form .) We claim that a WZW model on G WZW model is nearly identical to the action of G itself,
admits an action of the 2-group14 the difference being that the former includes an addi-
tional B field transformation. The reason such a similar
1 −→ BU(1) −→ G̃1 −→ G −→ 1, structure is possible is mathematically that the gerbe ad-
mits a canonical G-equivariant structure, and so G has
13 In a nonlinear sigma model on a space X , a closed B field de- a natural action on the gerbe defining the WZW model.
That said, the B field transformations induce subtle dif-
fines an analogue of a θ angle, twisting nonperturbative sec-
tors by phases given by the holonomy of the B field on various ferences in phases which manifest elsewhere. For exam-
2-cycles, which is trivial only when B is exact. Here, however, ple, if we apply the same reasoning to states defined on
we assume G is simple and simply-connected, in which case
both H 2 (G, Z) and H2 (G, Z) vanish.
14 Because we want to preserve the B field defined intrinsically by 15 Mathematically, the 2-group action on the gerbe is described in
the Wess-Zumino term, we only consider 2-groups defined by [28][section 7.3], and its multiplication in terms of Wilson lines
extensions by BU(1), and not BU(1)conn . (i.e. on the loop space) was described explicitly in section 2.
Fortschritte
der Physik
Fortschr. Phys. 63, No. 11–12 (2015)
Progress
Original Paper
of Physics
hemispheres, then the B field transformations generate looping of the 2-group G̃1 . Phrased differently, the de-
phase rotations on states, which the G action by itself though not a group, is the 2-group G̃1 .
looping of LG,
would not provide. Thus, in principle states in canonical So far we have briefly reviewed how the 2-group
quantization detect the difference between the actions of G̃1 has a natural action on the WZW model by com-
G and G̃1 , via differences in phases. (Also, the B field ac- bining the G action with B field gauge transforma-
tions of the G̃1 multiplication also manifest in D-brane tions (in fact, the underlying U(1) gerbe is the one
actionas, as has been discussed in [77].) that defines the Wess-Zumino term), and that the Kac-
Now, left- and right-multiplication by elements of G is Moody algebra is the infinitesimal algebra of that same
a global symmetry (though not a local symmetry) of the 2-group.
classical theory. However, in the WZW model, the algebra The reader might ask at this point about the Noether
of their left- and right-moving Noether currents J a has current associated with this 2-group G̃1 and its left-
a central extension, which in a dual theory of fermions and right-multiplications. Since the action of G̃1 on the
would correspond to a two-point anomaly. As is well- fields of the WZW model is, in the closed string the-
known, the resulting symmetry algebra is a Kac-Moody ory, effectively indistinguishable from the action of G,
algebra, of the form the Noether currents for G̃1 should be the same as that
1
for G – namely, the Kac-Moody currents above. Put an-
Jna , Jm
b
= f abc Jn+m
c
+ knδ ab δn+m,0 , (1) other way, we propose that the Kac-Moody currents have
2
two interpretations: one interpretation as the Noether
where the f abc are the structure constants of the (finite- current of G (with a central extension in the algebra),
dimensional) Lie algebra. As a result of the central exten- and another interpretation as the Noether current of
sion, the currents J a do not transform in a representation G̃1 (in which we propose that the central term be in-
of the group G, unlike primary fields, but rather trans- terpreted in terms of the B field transformations in G̃1 ,
form under the group action as [78][equ’n (3.17)] reflecting the phases picked up by states in canonical
quantization).
1 As evidence for the interpretation above, we ob-
δω J(z) = [ω(z), J(z)] + kω (z)
2 serve that a Noether current for the B field transfor-
for ω a Lie-algebra-valued function. mations should be a two-form current, which would
Those Kac-Moody algebras have a natural interpre- dualize in two dimensions to a scalar. We propose there-
tation in terms of the 2-group G̃1 , as infinitesimal auto- fore that the central extension term in the Kac-Moody
morphisms, as has been discussed in e.g. [8]. Now, Kac- algebra is the Hodge dual of that two-form current, and
Moody algebras are perhaps more widely understood in that the entire Kac-Moody algebra be understood as the
terms of the algebra of an extension of the algebra of the Noether current for G̃1 multiplication. (The fact that
loop group LG of G, but these are closely related. Specifi- the central extension is dually described in terms of an
cally, the looping of the 2-group G̃1 , anomaly, will be interpreted later in this paper as an ex-
ample of anomalies transmuting a classical group sym-
1 −→ BU(1) −→ G̃1 −→ G −→ 1 metry into a higher group symmetry of the quantum
theory.)
of the
is precisely16 the analogous central extension LG To close an intellectual loop, it is natural to con-
loop group LG, jecture that, for suitable definitions, representations of
−→ LG −→ 1. the 2-group G̃1 should coincide with representations
1 −→ U(1) −→ LG
of the corresponding Kac-Moody algebra, and indeed
(This is a special case of the statement that a U(1) this conjecture has been made by others (see e.g. [79]
1-gerbe over a space X lifts to a principal U(1) bundle [appendix A], [26]). After all, the states of a WZW model
over the loop space LX .) In this fashion, we can under- form representations of a Kac-Moody algebra, and we
stand the more historically common interpretation of the have also argued that the 2-group G̃ defines a global sym-
Kac-Moody algebra, in terms of extensions of LG, as a metry of the theory. Unfortunately, we are under the im-
pression that the representation theory of 2-groups is
not sufficiently well-developed to address this issue. (For
16 See [8, 21, 22] for further discussion of the relationship between current work on representations of 2-groups, see for ex-
loop groups and 2-groups. It should be noted that for this dis- ample [24, 80, 81].)
cussion, we specifically need G̃1 and not G̃c , as the latter is not In this language, it is tempting to speculate that
so closely related to the loop group extension. level-rank duality should be realized as some sort of

Fortschritte
der Physik
Progress
Original Paper
of Physics
equivalence, perhaps a Morita equivalence [82], of As such examples have been discussed extensively else-
2-groups, for example as where, and will play no further role in this paper, we will
not discuss them further here.
n,
k = SU(k)
SU(n)
where subscripts indicate levels. 3.5 Current algebras

Let us conclude this section by observing certain for-
mal similarities between the structure in this section and We have discussed how Kac-Moody currents in two-
our discussion of Dijkgraaf-Witten boundaries in sec- dimensional WZW models can be understood as Noether
tion 3.3. There, given a bulk Dijkgraaf-Witten theory in currents for a symmetry 2-group that mixes an ordinary
three dimensions, corresponding to a finite G gauge the- group with BU(1).
ory, we argued that any boundary theory should have a In this section we shall conjecture an analogous role
G̃ symmetry, where for current algebras in four dimensions.
In general, for two currents Jμ,α , Jμ,β (μ a spacetime
1 −→ BU(1) −→ G̃ −→ G −→ 1 Lorentz index), their equal-time commutation relations
will have the form [89][section 22.6], [90][section 11-3]
is the extension determined by the same element of
H 3 (G, U(1)) that determined the bulk theory. In this
x, t), Ji,β (
[J0,α ( y , t)] = Cαβγ Ji,γ ( x − y )
x, t)δ d−1 (
section, we have studied WZW models. However, WZW
models can be understood as boundaries of three- x − y ),
+ Sαβ,i,j ∂j δ d−1 (
dimensional Chern-Simons theories (see e.g. [83]). The
Chern-Simons theory is a G gauge theory, classified by an where the second term on the right-hand-side is a
element of H 3 (G, Z) (the level), and we have argued that c-number term known as the Schwinger term, and we
the boundary WZW theory admits a G̃ symmetry, where have formally written the expression for general dimen-
sions d, instead of specializing to d = 4.
1 −→ BU(1) −→ G̃ −→ G −→ 1 The Schwinger terms above have long been inter-
preted as defining a projectivization of the algebra of
is the extension determined by the same element of maps from the spacetime into a classical Lie algebra g
H 3 (G, Z) that determined the bulk theory. [91–94], and we propose a refinement of this idea. Specif-
In the case of discrete torsion in two dimensions ically, we propose that the algebraic structure above be
and Dijkgraaf-Witten theory in three dimensions, it was understood as a higher group extension of the form
essential that the boundary theory couple to the bulk
theory in such a way that the boundary form potential 1 −→ Bd−1 U(1) −→ G̃ −→ G −→ 1
transform when the bulk gauge field undergoes a gauge
transformation. Thus, the parallel would be especially generalizing the structure of Kac-Moody algebras. (For
meaningful if WZW models and Chern-Simons theories example, if the four-dimensional spacetime is R × T 3 ,
coupled in an analogous fashion. The reference [83] re- then a projectivization of maps into g is the Lie algebra
stricts to bulk gauge transformations that are trivial along of a bundle on the triple loop group L3 G, and the triple
the boundary, but if we consider more general cases, it loop space of the higher group above is exactly a bundle
is straightforward to see that under a gauge transforma- on L3 G.) The current for the Bd−1 U(1) piece is a d-form,
tion, the Chern-Simons action picks up an integral of a which dualizes in d dimensions to a scalar. We conjecture
total derivative. In principle one could imagine that this that scalar should be interpreted as the Schwinger term.
could be cancelled by a symmetry transformation of the As a consistency check, note that extensions of the form
WZW B field. above are classified by homotopy classes of maps17
In passing, we should mention that such
bulk/boundary relationships have been studied more [G, BBd−1 U(1)] = [G, Bd U(1)]
generally in e.g. [2, 84, 85] in extended TQFT’s, and also
= [G, K (Z, d + 1)] = H d+1 (G, Z)
in related work [86–88]. Roughly speaking, the form of
the results is that a Bk G gauge theory in d dimensions
has on its boundary a theory with symmetry G̃ given as
the extension 17 For readers unfamiliar with the notation, [X, Y ] for X , Y
spaces or stacks denotes the homotopy classes of maps from X
1 −→ Bd−2 U(1) −→ G̃ −→ Bk G −→ 1. to Y .
Fortschritte
der Physik
Fortschr. Phys. 63, No. 11–12 (2015)
Progress
Original Paper
of Physics
as discussed in appendix A. Indeed, in higher- r π1 (M) counts possible cosmic strings, determined by
dimensional analogues of WZW models as discussed winding in the scalar vevs,
in18 e.g. [96], the analogue of the Wess-Zumino term r π2 (M) counts analogues of monopoles, here deter-
in d dimensions is defined by a degree (d + 1) form, mined by noncontractible two-spheres in M, and
integrated over a bounding (d + 1)-dimensional space, r π3 (M) counts textures.
which is precisely consistent with the classification
Examples often arise in gauge theories. Given a G
above.
gauge theory with a Higgs field that breaks G to H ⊂ G,
and a suitable nontrivial potential19 , the moduli space M
of Higgs vevs can be taken to be the coset G/H. The ho-
3.6 Other Kac-Moody actions
motopy groups of M = G/H can be related to the homo-
topy groups of G and H using the homotopy long exact
We have seen that Kac-Moody algebras are closely re-
sequence, for example
lated to Lie algebras of 2-groups, and so it is worth point-
ing out that there are other actions of Kac-Moody alge- πk (G) −→ πk (G/H) −→ πk−1 (H) −→ πk−1 (G).
bras in field theories in dimensions greater than two. We
list a few examples below. We shall see applications of these ideas momentarily.
The work [97–100] described Kac-Moody actions Now, let us outline a classification of analogues of
on instanton moduli spaces (see also [101]), albeit at topological defects pertinent to one-form symmetries
level 0. More recently, other Kac-Moody actions on four- and corresponding two-groups. We claim that the rel-
dimensional gauge theories have been discussed in e.g. evant analogue of the moduli space20 M of scalar field
[102, 103], though again at level 0. As discussed in vevs is the loop space LM parametrizing Wilson lines.
[104][section 4], actions of Kac-Moody algebras on co- After all, Wilson line vevs are sections of a bundle over
homology of moduli spaces of instantons on ALE spaces LM, a bundle determined by the BG action, in the same
were discussed in e.g. [105]. way that for ordinary symmetries, the scalars are sections
Examples of hyperbolic Kac-Moody algebra actions of a bundle over M. More explicitly, (low-energy) Wilson
on supergravity theories are discussed in a number of ref- lines depend upon scalars explicitly as, for example,
erences including e.g. [106–110].
P exp φ a Aaμ dsμ ,
4 Cosmological defects and generalized

hence a given Wilson line is determined by a loop in
moduli spaces M, in line with observations above about cosmic strings,
and becomes a (not necessarily neutral) scalar over the
4.1 Generalities on defects and ordinary moduli spaces (based) loop space LM.
Then, just as homotopy of the space of scalar vevs M
In four dimensional theories, there are a variety of cos- can encode information about topological defects perti-
mological defects that are typically classified by the ho- nent to an ordinary group G, homotopy of the loop space
motopy of the moduli space of scalar vacuum expecta-
tion values. For example, if M denotes that moduli space,
then [111][section 3.2]: 19 We are interested in topological defects, not semilocal strings.
r π0 (M) counts components of M, hence counts pos-
In the latter, the potential could vanish identically, so the full
moduli space of Higgs vevs also includes e.g. vanishing Higgs
sible domain walls,
vevs, or other vevs for which the unbroken subgroup is different
from that above.
20 Or stack. See [55] for a discussion of cases in which the moduli
18 See also [95] for an updated description relevant for more gen- ‘space’ admits a stack or gerbe structure, and in particular the
eral 4-manifolds. The analysis there interprets these structures relevance to cosmic strings and so forth of homotopy groups
in terms of differential cohomology theories. A thorough de- of the resulting gerbe. We should add that, although that refer-
scription of current algebras along the lines we suggest should ence did not make a clear statement regarding existence of cos-
take that into account, but as we are only outlining a conjec- mic strings and so forth associated with trivially-acting finite
ture, we shall not try to relate those differential cohomologies groups, we have since come to believe that they do exist, hence
to higher groups in this brief section. Instead, we leave such an we now believe that homotopy groups of Deligne-Mumford
analysis for future work. stacks should indeed have physical relevance.

Fortschritte
der Physik
Progress
Original Paper
of Physics
the Wilson line to be determined in part by the cos-

mic string. In other words, the gauge field along the
Wilson line should couple to a bundle on C induced
by the cosmic string (as a ’t Hooft line). Now, it is not a
coincidence that, at least for G connected and simply-
connected, the classification of cosmic strings in this
theory matches the classification of induced bundles
on circles. Specifically, from the homotopy long exact
Figure 2 A Wilson line about a closed loop itself moves through a sequence, if G is connected and simply-connected,
closed path, which encloses a monopole (right side). π1 (G/H) = π0 (H). Now, principal H bundles on S1 are
classified by
LM of scalar field vevs can encode information about [S1 , BH] = π1 (BH) = π0 (H),
topological defects pertinent to BG. To that end, there is
a key identity relating the homotopy of M and its based so we see that the classification of cosmic strings
loop space LM: matches that of principal H bundles on a circle, as
one would expect from the fact that the cosmic string
πk (M) ∼
= πk−1 (LM). (as a ’t Hooft loop) is inducing a bundle on the enclos-
ing circle. In particular, a cosmic string acts as a do-
Alternatively, the reader might prefer to think about
main wall in the space of Wilson line vevs: depending
analogues of Wilson lines for BG symmetries as defined
upon whether the curve C encloses a cosmic string
by Wilson surfaces of the form
(and the number of times it wraps), one should get a
different component of the space of Wilson line vevs.
exp F r Analogues of cosmic strings for Wilson lines should

in principle be counted by π1 (LM) = π2 (M), which
for some two-dimensional submanifold and F a two- correspond to monopoles. To help explain why this
form. Broadly speaking, these would be associated with is sensible, first recall that a cosmic string can be in-
elements of π2 (M), but using the identity above, terpreted as a ’t Hooft line [112]. A charged particle
that walks along a loop enclosing the ’t Hooft line
π2 (M) ∼
= π1 (LM) will pass through a transition function for a nontrivial
bundle generated by the ’t Hooft line. The analogue
and so again we can reduce these considerations to prop-
of a ’t Hooft line for a BG symmetry is a monopole,
erties of the loop space.
which generates a nontrivial G-gerbe. Indeed, as one
Let us walk through the implications of this statement
drops a monopole through a Wilson line, its vev will
for Wilson lines for one-form symmetries:
change21 , so in this fashion one has a notion of wind-
r Analogues of domain walls for Wilson lines should in ing for Wilson lines. A Wilson line walking through a
principle be counted by π0 (LM) = π1 (M), which cor- loop in which it wraps a monopole is schematically
respond to cosmic strings. Here the intuition is sim- illustrated in figure 2.
ply that, as the cosmic string corresponds to a ’t Hooft As a particular example, let us consider analogues of
loop dual to the Wilson line [112], then the Wilson line Alice strings. Recall (see e.g. [111][section 4.2.4], [112–
changes as it winds around the cosmic string. Thus, in 115]), Alice strings arise in SO(3) (and SU(2)) gauge
this sense, cosmic strings can define analogues of do- theories with matter in the 5-dimensional representa-
main walls for Wilson lines. tion, consisting of real symmetric traceless 3 × 3 ma-
For a simple example, consider the case of a G gauge trices. A typical Higgs vev, e.g.
theory broken to H ⊂ G by a Higgs vev, as outlined ⎡ ⎤
above. The cosmic strings are classified by π1 (G/H). 10 0
⎣0 1 0 ⎦
Now, if we run a Wilson line
0 0 −2

P exp A · ds
C
21 See for example [28][section 7.1], but in fact at some level this is
around a cosmic string, by taking the curve C to en- merely an unwinding of the definition of a monopole across the
close the cosmic string, then we expect the phase of surface of an enclosing S2 .
Fortschritte
der Physik
Fortschr. Phys. 63, No. 11–12 (2015)
Progress
Original Paper
of Physics
has stabilizer S(O(2) × O(1)) = O(2) ⊂ SO(3), so with that

a suitable potential forcing the vacua to lie on the
coset, the moduli space of Higgs vevs is SO(3)/O(2) = π2 (SU(3)/SO(3)) = Z2 .
RP2 . (Equivalently, we could work in an SU(2)
gauge theory, in which case the stabilizer of the In particular, a Wilson line following a path that
Higgs vev above would be22 Pin(2) ⊂ SU(2), and again swoops around a monopole, as in figure 2, would un-
SU(2)/Pin(2) = RP2 .) This moduli space has π1 = Z2 , dergo a gauge transformation in the nontrivial ele-
hence there are cosmic strings in this model, in which ment of Z2 .
the Higgs field vev wraps nontrivially around the
r Analogues of monopoles for Wilson lines should in
moduli space. More to the point, the cosmic strings principle be counted by π2 (LM) = π3 (M), which cor-
are classified by Z2 . The Wilson lines act on electric respond to textures.
charges of the unbroken O(2) (Pin(2) in the SU(2)
r Finally, analogues of textures for Wilson lines should
theory) by flipping their signs: representations of in principle be counted by π3 (LM) = π4 (M).
O(2) (resp. Pin(2)) are pairs of matched SO(2) (resp.
Another standard matter involves defects ending on
Spin(2)) representations of opposite sign [116][sec-
other defects. For example, suppose a gauge symmetry G
tion 11.1], and the Wilson line about the string, which
is broken in two steps at successively lower energies, first
takes values in the diconnected component, flips the
to a subgroup H ⊂ G, and then H is completely broken.
signs, exchanging the two SO(2) (resp. Spin(2)) rep-
(See e.g. [40][section 5.7.1] for a longer review.)
resentations. The fact that electric charges undergo
a sign flip under parallel transport about the string, r Domain walls ending on strings: After the first step,
ultimately a consequence of the fact that the strings when G is broken to H, the theory will have cosmic
are classified by Z2 , is the defining characteristic23 of strings counted by π1 (G/H). After the second step,
Alice strings. the theory will have domain walls counted by π0 (H).
The analogue of Alice strings for Wilson lines would The domain walls of the second step can end on the
involve a moduli space M such that π1 (LM) = strings of the first step, consistent with the observa-
π2 (M) = Z2 . In general, for a G gauge theory with tion that (for G connected and simply-connected),
a potential forcing all the Higgs vevs to break G to π1 (G/H) ∼
= π0 (H).
the same subgroup H ⊂ G, the moduli space of Higgs r Strings ending on monopoles: After the first step,
vevs is the coset G/H. For example24 , consider an when G is broken to H, the theory will have
SU(3) gauge theory in which a Higgs vev breaks the monopoles counted by π2 (G/H). After the second
SU(3) to SO(3) ⊂ SU(3). This would happen for a step, the theory will have strings counted by π1 (H).
Higgs field in the Sym2 3 = 6 representation, with a The strings of the second step can end on the
vev given by the identity. This theory has the property monopoles of the first step, consistent with the ob-
servation that (for G 1- and 2-connected), π2 (G/H) ∼=
π1 (H).
The analogue for Wilson lines is similar:

22 Pin(2) here can be understood as the normalizer of a maximal
torus. Pin(2) is also a double-cover of O(2).
r Domain walls ending on strings: After the first step,
23 As an aside for interested readers, let us discuss another char- when G is broken to H, the theory will have cos-
acteristic feature of Alice strings in this language. Take a SO(2) mic strings counted by π1 (LG/H) = π2 (G/H). After
charge q, and pass through the middle of a pair of parallel the second step, the theory will have domain walls
Alice strings. Asymptotically, the pair of parallel Alice strings counted by π0 (LH) = π1 (H). Thus, the analogue of
yield an O(2) bundle whose disconnected component is trivi- domain walls ending on strings, can equivalently be
alizable, so SO(2) charges are sensible asymptotically (though understood as ordinary strings ending on ordinary
not invariant under O(2) gauge transformations). Pass a charge monopoles.
q through the middle of the pair, and it will come out on the
other side as charge −q, hence to preserve charges, the pair of In principle, the same analysis can be continued for
Alice strings are then interpreted as having charge +2q. This higher p-form symmetries. For example, analogues of
phenomenon is known as ‘Cheshire charge.’ The basic issue in cosmological defects for a Wilson surface corresponding
all these cases seems to be one of ill-fated attempts to define to a 2-form symmetry would be measured by πk (L2 M) =
SO(2) matter in an O(2) gauge theory. πk+2 (M). For example, a domain wall for a Wilson sur-
24 We would like to thank A. Knutson for suggesting this example. face would correspond to an ordinary monopole: the set

Fortschritte
der Physik
Progress
Original Paper
of Physics
of values for a Wilson surface would break into compo- rather than πm (M). In the case above, the theory with
nents, determined by how many times the Wilson sur- k = 1 admits only monopoles, as only π2 (M) = 0. The
face wraps the monopole. corresponding theory with k > 1 also admits monopoles,
In a similar fashion, analogues of cosmological de- but the counting is slightly different. In particular, since
fects for a Wilson surface corresponding to a p-form sym- π2 (BZk ) vanishes, the homotopy sequence above reduces
metry would be measured by πk (Lp M) = πk+p (M). Of to
course, for a spacetime of a given dimension, there is an
upper limit to the degree of homotopy groups that can be 0 −→ π2 (M) −→ π2 (M)(∼
= Z) −→ π1 (BZk )(∼
= Zk ) −→ 0.
realized in that spacetime. The intuition behind higher
Thus, our theory with k > 1 also admits countably many
examples rapidly becomes obscure, so we will not pur-
monopoles, but counted slightly differently than in the
sue this further, except to note that this does provide a
case k = 1. In particular, we see that not every monopole
physical interpretation to higher homotopy groups of the
in the k = 1 theory lifts to a monopole in the k > 1 theory
moduli space M of scalar field vevs.
– apparently only monopoles whose charges satisfy a di-
visibility constraint lift to the k > 1 theory, as one might
4.2 Generalized moduli spaces
expect on general principles of charge quantization.
As one more example, consider a four-dimensional
So far we have discussed cosmological defects and higher
theory in which one gauges a finite group G, all of which
loops for ordinary moduli spaces. Next, we shall consider
acts trivially on the theory. If the ungauged theory had
cases in which a higher group action exists on the moduli
moduli space M, then the gauge theory has moduli
space of the field or string theory. In examples in which
‘space’ M × BG. In this case, the counting is again modi-
the moduli ‘space’ admits an action of a higher group, the
fied, but in a more trivial fashion, as
‘space’ is a generalized space known as a stack.
The paper [55] studied a number of properties of π1 (M × BG) = π1 (M) ⊕ π1 (BG)
four-dimensional theories whose moduli ‘spaces’ were
of this form. For completeness, we begin by reviewing and other homotopy groups are unmodified. Again the
the highlights of cosmological defects studied in [55], counting of cosmological defects is modified, but in a
here. We will then turn to other moduli ‘spaces’ appear- much more trivial fashion.
ing in field and string theory, motivated by our discus- So far we have considered moduli ‘spaces’ of four-
sion of higher group actions on WZW models and current dimensional field and string theories admitting an ac-
algebras. tion of BG for G finite. Mathematically, there is a notion
To begin, consider as a prototypical example a four- of generalized ‘spaces’ that would admit actions of BG
dimensional U(1) gauge theory with 2 chiral multiplets of for G nonfinite; however, it is not known at present how
charge +k and 2 chiral multiplets of charge −k, for k > 1, to, for example, define a sigma model on such. (Tech-
along with massive minimally-charged matter. As dis- nically, it is known how to define a sigma model on a
cussed earlier in section 3.2 as well as e.g. [50–52, 55, 56], Deligne-Mumford stack, but not an Artin stack.) In the
both the quantum field theory and the moduli ‘space’ ad- remainder of this section, we will explore the possibility
mit an action of BZk , and so its spectrum of cosmological of whether such more general moduli ‘spaces’ might ap-
defects is slightly different than for ordinary cases. pear in physics via a simple example. We will get rather
Let M denote the moduli ‘space’ of the theory above, confusing results, suggesting that possibly the only gen-
and M the ordinary moduli space of the corresponding eralized moduli ‘spaces’ relevant for physics are of the
theory with k = 1. As discussed in [55], there is a long ex- former (Deligne-Mumford) type.
act sequence of homotopy groups
ordinary cosmic strings, in other words that π1 (BZk ) is count-
· · · −→ πm (BZk ) −→ πm (M)
ing cosmic strings of this type. At the time reference [55] was
−→ πm (M) −→ πm−1 (BZk ) −→ · · · . written, this matter was ambiguous, but now seems to have
been settled in the affirmative. Such cosmic strings are defined
In principle, one would expect cosmological defects by nontrivial Zk bundles over enclosing circles, but in which
in the theory with k > 1 to be counted by25 πm (M), the Zk acts trivially on the other fields of the theory. (In particu-
lar, a map from any space X into BZk is defined by a Zk bundle
on X , both for the classifying space and for the stack, which is
25 This assumes that e.g. cosmic strings defined by nontrivial part of the reason why cosmic strings on such higher groups
bundles for trivially-acting groups exist and are different from would have the form described.)
Fortschritte
der Physik
Fortschr. Phys. 63, No. 11–12 (2015)
Progress
Original Paper
of Physics
In section 3.4 we reviewed existing results on how Unfortunately, it is not clear how to interpret the ho-
WZW models are believed to admit an action of G̃, an motopy groups of the Artin stack above physically. It is
extension of a Lie group G (partially defining the WZW possible that one must perform a more subtle analysis
model) by BU(1), where the extension class is deter- along the lines of [95]. However, for the moment we will
mined by the level of the WZW model. We could imagine instead interpret this to mean that the physical relevance
fibering the WZW model over some other space X , as de- of Artin stacks is unclear, and when computing e.g. cos-
scribed in e.g. [118–123]. The resulting two-dimensional mological defects, one should restrict to cases involving
theory has a semiclassical moduli ‘space’ given by a Deligne-Mumford stacks (meaning, actions of BG for G
G̃-bundle over X , another example of a stacky moduli finite).
space, albeit an Artin stack and not a Deligne-Mumford
stack.
The two-dimensional example above would, for obvi- 4.3 Analogues of Goldstone’s theorem and generalized
ous reasons, not admit a wide variety of cosmological de- moduli spaces
fects, but formally we can imagine higher-dimensional
analogues. Following the proposal in section 3.5 for an Previously in section 3.1 and in [1], analogues of Gold-
interpretation of current algebras in terms of higher stone’s theorem were discussed for counting Goldstone
groups, one could similarly imagine fibering a higher- bosons in theories with spontaneously broken higher
dimensional current algebra over some space X , at least group symmetries. For example, the Goldstone boson as-
in low-energy effective field theory. If G̃ is the higher sym- sociated with a spontaneously broken B U(1)conn in an
metry group of the current algebra, then by analogy with abelian gauge theory was proposed in [1] to be the gauge
fibered two-dimensional cases, one would have a moduli field itself.
‘space’ given by a G̃-bundle over X . Now, for ordinary groups, there is additional content
In such a case, since G̃ was determined by an exten- in Goldstone’s theorem. In principle, Goldstone’s theo-
sion encoding an anomaly, in effect some of the homo- rem not only gives a count of bosons associated with
topy groups of G̃ and hence of the moduli stack would be a spontaneously broken global symmetry, but also de-
determined by an anomaly. Put another way, one would scribes the local dimension of the moduli space. After
have anomaly-induced cosmological defects. all, if a global symmetry G is broken to a subgroup H,
For example, let us consider the SU(3) current algebra then one has dim G/H massless bosons ultimately be-
described in [96]. In present language, section 3.5 conjec- cause the moduli space of vacua locally looks like G/H.
tures that it should have a G̃ symmetry, where G̃ is given We have already discussed how generalized moduli
by ‘spaces’ (stacks) can admit actions of higher groups, so it
is natural to ask what analogous statements can be made
1 −→ B3 U(1) −→ G̃ −→ SU(3) −→ 1, for the moduli stacks appearing in such gauge theories.
For example, if one spontaneously breaks a BG symmetry
where the level k of the current algebra (n in the notation to a subgroup BH, then does a coset of the form BG/BH
of [96]) determines the extension class. We can consider play a role in physics, in the same way that for ordinary
this model by itself, in effect fibering over a point. groups, cosets G/H play a role in understanding moduli
From the long exact homotopy sequence (which op- spaces in spontaneous symmetry breaking? In this sec-
erates for higher groups in essentially the same way tion, we will not reach any firm conclusions, but we will
as for spaces [27]), we see that for m = 4, 5, πm (G̃) = discuss these issues in simple examples.
πm (SU(3)), which (for n = 4, 5) is nonzero for n = 3 and Let us first outline a little pertinent mathematics. If
n = 7 for example. For m = 4, 5, there is a long exact G̃ is a higher group, and H̃ a sub(higher)group, then the
sequence homotopy quotient G̃/H̃ exists [27], and there is a long
exact sequence of homotopy groups of exactly the same
0 −→ π5 (G̃) −→ π5 (SU(3))(∼
= Z) −→ π4 (B3 U(1))(∼
= Z) form as for ordinary groups [27], namely
−→ π4 (G̃) −→ 0.
πk (G̃) −→ πk (G̃/H̃) −→ πk−1 (H̃) −→ πk−1 (G̃).
The map π5 (SU(3)) → π4 (B3 U(1)) should be propor-

tional to the level k of the current algebra, so that at level For the higher groups Bk G, we can use the fact that
zero, the sequence splits, and π5 (G̃) ∼ = π5 (SU(3)). If the
level k = 0, then π5 (G̃) = 0 and π4 (G̃) = Zk . πm (Bk G) = ππm−1 (Bk−1 G) = · · · = πm−k (G)

Fortschritte
der Physik
Progress
Original Paper
of Physics
to relate the homotopy groups of Bk G to those of G. naive application of a generalized Goldstone’s theo-
The ‘spaces’ of such higher groups can be understood as rem as giving BU(1) directions in the moduli ‘space’
stacks, or ‘generalized spaces’ in the language of the pre- in the spontaneously broken symmetry phase. That
vious section. said, very naively, similar statements could also be
Second, it may be pertinent to distinguish two cases. made in other dimensions, even in which the large-
As previously outlined, stacks associated with BG for G loop behavior changes and one no longer has an
finite are different both mathematically and physically IR interpretation in terms of spontaneous symmetry
from those associated with BG for G non-finite. For the breaking.
former, known technically as Deligne-Mumford stacks, it
is known how to define a sigma model [52]. For the latter, At this point in time, we are therefore not able to make
known technically as Artin stacks, the physics is largely any firm statements.
unknown, and we suspect likely to be more subtle than
for Deligne-Mumford stacks. As a result, it is entirely pos-
sible that physical interpretations of homotopy groups of 5 Anomalies as transmutation
gerbe structures may differ between such cases.
Now, let us consider a concrete example. As was re- 5.1 Two-dimensional cases
viewed in section 3.1, reference [1] discusses abelian
gauge theories as examples of theories with B U(1)conn Let us now turn to the application of 2-groups to anoma-
actions. As discussed there, in some dimensions, this lies in two-dimensional quantum field theories. If J is a
symmetry is spontaneously broken, and the abelian Noether current, or rather the corresponding composite
gauge field is identified with the Goldstone boson. Let operator in some two-dimensional quantum field theory,
us try to interpret this result in terms of possible moduli then the J 2 anomaly in two dimensions is well-known
stack structures. to be proportional to Tr F ∧ F, involving the curvature of
a coupled (possibly nondynamical) gauge field. Anoma-
r First, let us proceed naively, and look for an inter- lies of this sort are indicative of curvature on the space
pretation of the homotopy groups of B U(1)conn itself. of gauge transformations, or somewhat more precisely,
Now, as outlined in section 2, the homotopy groups of indicate that the path integral measure fails to be well-
B U(1)conn are the same as those of K (U(1), 0), which defined under families of gauge transformations. (See for
is to say, example [124][section 3] for a pertinent discussion.) This
can happen even if there are no chiral fermi zero modes:
π0 (B U(1)conn ) = U(1), the path integral measure can be invariant under a single
πn (B U(1)conn ) = 0 for n > 0. gauge transformation, but not well-behaved in families.
The fact that the anomaly arises in families of
However, it is not entirely clear to us how these homo- gauge transformations suggests an interpretation of the
topy groups would be interpreted in an abelian gauge anomaly. Although, the anomaly ‘breaks’ the classical
theory. symmetry group G, the form in which it is broken sug-
r Next, let us take a slightly more radical approach. gests that the quantum theory might have a different
Since we have an abelian gauge theory, it is tempt- symmetry ‘group,’ perhaps one in which associativity is
ing to replace the moduli space of the theory M with broken up to higher cocycles, where the associators are
an Artin stack, a U(1) gerbe over M with fibers BU(1). encoded implicitly in the phases one picks up around
This would be in the spirit, though not the detailed loops.
method, of [50, 52, 55]: a sigma model on a stack In this section, we will propose that in such cases, the
of this form should morally involve a path integral quantum theory has a symmetry 2-group, instead of a
over U(1) bundles with connection, hence an abelian symmetry group, specifically the 2-group G̃1 defined in
gauge theory, though previous work has only made [6] by an extension
claims about stacks involving finite gerbes (Deligne-
Mumford stacks). Sigma models on Artin stacks have 1 −→ BU(1) −→ G̃ −→ G −→ 1.
not been studied at all, and are likely to be subtle to
interpret. That said, if this extremely naive approach In particular, in this interpretation, the anomaly breaks
were to hold water, then in principle we could ef- the classical symmetry, but simultaneously encodes an
fectively encode the U(1) gauge fields in the ‘stacky’ extension of G to a 2-group symmetry, and so describes a
structure of the moduli space, consistent with the transmutation of a zero-form symmetry into a 2-group
Fortschritte
der Physik
Fortschr. Phys. 63, No. 11–12 (2015)
Progress
Original Paper
of Physics
symmetry, a structure we have seen elsewhere in this is described by a principal G bundle P over the target
paper. space. Given a principal G bundle P, there is an obstruc-
One way to see this structure is via a dual representation to lifting it to a principal G̃1 bundle. For the same
tion of the fermions, in a WZW model. As is well-known reasons as above, that obstruction is
[76], the nonabelian bosonization of fermions with clas-
sical symmetry group G is described by a WZW model, k Tr F ∧ F,
and we described previously in section 3.4 how WZW
where k ∈ H 3 (G, Z) is the level of the WZW model, or
models form examples of theories in which ordinary
equivalently the cohomology class describing G̃1 as an
group actions are replaced with 2-group actions.
extension of G. (For example, for a trivial extension, k =
In this language, the fermion chiral symmetry cur-
0, and there is no obstruction.) In a heterotic theory,
rents correspond to, in the WZW model, classically the
this obstruction corresponds to a contribution to the
left and right multiplications by G. These classical sym-
Green-Schwarz anomaly in a (0,2) theory describing a
metries are modified in the quantum theory by the cen-
left-moving current algebra at level k, for example.
tral extension term in the Kac-Moody algebra, which,
Consider more general fibered WZW models, in
as has been argued here and elsewhere, suggests they
which one bosonizes both left and right-movers. Sup-
should be extended to left and right multiplication by the
pose one has classically a principal GL × GR bundle, with
2-group G̃1 , defined by the extension of G by BU(1) dic-
different groups for left and right-movers. If we call this
tated by the level of the WZW model. The fact that the
bundle P, then anomaly cancellation in this case be-
2-group G̃1 explicitly appears in the bosonized represen-
comes the statement that the GL × GR bundle can be
tation of the fermions, argues that it must also appear
lifted to a G
L × GR1 bundle; obstructions on either side
in the quantized fermion theory, and since in the WZW
separately are cancelled out when one takes both factors
model the 2-group G̃1 extends the classical group G, in
together.
the dual fermion theory the 2-group G̃1 must replace the
The idea that a classical group can transmute into a
anomalous action of the classical symmetry group G.
2-group symmetry of the quantum theory, as we have
Another check of this proposal arises when we couple
formulated it above, may seem odd, but analogous trans-
the fermions to a gauge field. In principle, if the classical
mutations may also sometimes appear along renormal-
symmetry group becomes in the quantum theory a sym-
ization group flows. For example, in two dimensions it is
metry 2-group, then to gauge G, any classical principal G
a common result that global symmetry groups become
bundle P must be lifted to a principal G̃1 bundle. The ob-
enhanced at IR fixed points to affine algebras, which as
struction to this lift26 is discussed in [6]; it is of the form
previously discussed are linked to 2-groups. In this lan-
k Tr F ∧ F, guage, at least naively RG flow would appear to describe
a transmutation from a (UV) global symmetry group to
where F is the curvature of P and k ∈ H 3 (G, Z) is the an (IR) 2-group.
level of the affine algebra, or equivalently the cohomol-
ogy class describing G̃1 as an extension of G. (For ex-
ample, for a trivial extension, k = 0, and there is no ob- 5.2 Conjectures on elliptic genera
struction.) Demanding that this obstruction vanish for
all bundles P is, in general, not possible, and so gauging We can gain further possible insight into the physical rel-
the classical symmetry is not possible, as one expects for evance of 2-groups by considering elliptic genera. The
an anomalous symmetry. basic Witten genus describes a quantum theory of a set
Another check arises in fibered WZW models [118– of fermions with a Spin rotation symmetry, fibered over
123], describing e.g. a bosonization of the fermions in a some space X , coupled via the tangent bundle.
nonlinear sigma model. The fibering of the WZW model In discussions of elliptic genera, it is well-known
that a topological group named String plays an impor-
tant role. For this discussion, we should point out that
26 Obstructions to spin structures may be more familiar to some
String(n) can be built as a 2-group, extending Spin(n) at
readers, and are analogous [117][section 3]. Given a principal level one by BU(1) [6][section 4.3], [7][section 7], [8]:
SO(n) bundle P over some manifold M, as is well-known there
is an obstruction to lifting to a principal Spin(n) bundle, mea- 1 −→ BU(1) −→ String(n) −→ Spin(n) −→ 1.
sured by the second Stiefel-Whitney class w2 (P). Correspond-
ing to this obstruction is a Z2 gerbe on M, whose characteristic With the above in mind, it is very tempting to con-
class is given by w2 (P) ∈ H 2 (M, Z2 ). jecture that the states appearing in elliptic genera form

Fortschritte
der Physik
Progress
Original Paper
of Physics
a representation of the 2-group String(n). In particular, an interpretation of certain anomalies in terms of higher
the states lie in representations of an affine extension of group symmetries of quantum theories.
so(n), and as we discussed in section 3.4, it is tempting There is also evidence for a role for 2-groups in moon-
to identify representations of the 2-group and the affine shine, see e.g. [81, 125][section 5]. It has also been sug-
algebra. However, as we discussed there, the representa- gested that such higher groups might play a further role
tion theory of 2-groups does not seem to be sufficiently in the understanding of Mathieu groups in SCFT’s as
well-developed at the moment to completely answer this described in e.g. [126–130], as for example the Mathieu
question one way or another, so we leave this matter for group M12 is a subgroup of a groupoid denoted M13 [26,
future work. 131].
In section 3.2, we discussed two-dimensional gauge
theories in which a finite subgroup G of the gauge group
5.3 Higher-dimensional conjectures acts trivially on the massless matter. We argued that at
low energies, these theories have a BG symmetry, and
In 2n dimensions, anomalies can arise from higher- outlined how these theories decompose into disjoint
dimensional analogues of the same issue with well- unions of simpler theories. It would be interesting to un-
definedness in families. As discussed in e.g. [124] derstand if there are analogues of decomposition for any
[section 3], consider a theory of charged fermions on S2n , notion of BG gauge theories or other higher gauge theo-
coupling to a principal G bundle P over some space X . ries in any dimension.
Under a one-parameter family of gauge transformations, It is tempting to speculate that some of the re-
the fermion path integral may fail to be well-defined, lations we have explored between ordinary groups
measured at least in part by an element of H 2n+1 (G). and higher groups may have analogues outlined in
Proceeding solely by analogy with the two- [132], but we shall leave such considerations for future
dimensional case discussed above, in our language, work.
this might correspond to the statement that the correct
symmetry group of the theory is a 2n-group, which Acknowledgements. We would like to thank M. Ando, N. Aramian,
topologically is a (2n − 1)-gerbe over G, an extension of N. Epa, D. Freed, N. Ganter, A. Knutson, T. Pantev, U. Schreiber,
G by B2n−1 U(1). If we take enough loops, this becomes and R. Thorngren for useful conversations. This paper began as
a central extension of L2n−1 G by U(1). It is natural an joint effort with M. Ando, who provided initial ideas and com-
to conjecture that obstructions to lifting principal G mentary, and N. Ganter and N. Epa similarly contributed to the
bundles to principal G̃1 bundles, for G̃1 a higher group development at an intermediate stage. U. Schreiber contributed
extension as above, are characterized by a degree 2n + 2 many references and results on higher groups, and N. Aramian also
characteristic class, of the form contributed many technical observations. Although we were mo-
tivated to write up our results by the publication of [1], this pa-
k Tr F ∧ · · · ∧ F per is based on discussions of applications of 2-groups to physics
that took place over several years, during which time E.S. was sup-
(n + 1 factors), which would certainly tie into the de- ported by NSF grants DMS-0705381, PHY-0755614, PHY-1068725, and
scription of anomalies, as well as central extensions in PHY-1417410.
higher dimensional current algebras as outlined in sec-
tion 3.5. We will leave this matter for future work.
A Topological classification of extensions
6 Conclusions In this paper we have encountered a number of higher
groups G̃ constructed as extensions of the form
In this paper we described the p-form symmetries of
[1] as special cases of higher group actions in quan- 1 −→ Bk U(1) −→ G̃1,k −→ G −→ 1,
tum field theory. After briefly outlining the definition of
1 −→ Bk U(1)conn −→ G̃c,k −→ G −→ 1,
higher groups, we outlined several examples, both re-
viewing some already in the literature as well as outlin- 1 −→ Bk U(1) −→ G̃0,k −→ G −→ 1,
ing some new examples. We discussed the role of ‘gener-
alized moduli spaces’ in quantum field theory, admitting for some other (higher) group G. To classify the possi-
actions of higher groups, and discussed cosmological de- ble extensions, we first need to compute the topological
fects in this context. We concluded with a proposal for classes of extensions.
Fortschritte
der Physik
Fortschr. Phys. 63, No. 11–12 (2015)
Progress
Original Paper
of Physics
An extension of the form above is the total space of a As a sanity check, note in the special case that k = 1,
Bk U(1) (or Bk U(1)conn , or Bk U(1)) bundle over G. Regard- for G of dimension greater than zero, the extensions are
less of our version of Bk U(1), for the purposes of a topo- classified by H 3 (G, Z), as consistent with WZW models
logical classification of bundles, we can replace it with a and Chern-Simons theories, for example. For k = 1, if G
suitable Eilenberg-Maclane space: is finite, the extensions are classified by H 3 (G, U(1)), con-
sistent with three-dimensional Dijkgraaf-Witten theory.
Bk U(1) = K (Z, k + 1), BU(1) ∼
= CP∞ .
(In particular, since the geometric realizations of all three

options are homotopic to one another27 , the resulting References
topological classification will not depend upon whether
we are extending by Bk U(1) or Bk U(1)conn or Bk U(1).) [1] D. Gaiotto, A. Kapustin, N. Seiberg, and B. Wil-
Over any space M, Bk U(1) = K (Z, k + 1) bundles are lett, “Generalized global symmetries,” arXiv:
classified by homotopy classes of maps into a next higher 1412.5148.
[2] D. Freed, “Higher algebraic structures and quan-
Eilenberg-Maclane space, i.e. homotopy classes of maps
tization,” Comm. Math. Phys. 159, 343-398 (1994),
hep-th/9212115.
M −→ K (Z, k + 2),
[3] J. Baez and J. Huerta, “An invitation to higher gauge
theory,” arXiv: 1003.4485.
which are computed by H k+2 (M, Z). Thus, we see that ex-
[4] H. Sati, “Geometric and topological structures re-
tensions of the form above are classified topologically by lated to M-branes,” arXiv: 1001.5020.
H k+2 (G, Z). [5] H. Sati, U. Schreiber, and J. Stasheff, “L∞ algebras
Now, in principle we are not yet done. We need to and applications to string- and Chern-Simons n-
classify extensions with higher group structures, not just transport,” arXiv: 0801.3480.
topological bundles. For 2-groups of the form [6] C. Schommer-Pries, “Central extensions of smooth
2-groups and a finite-dimensional string 2-group,”
1 −→ BU(1) −→ G̃1 −→ G −→ 1, Geom. Topol. 15, 609–676 (2011), arXiv: 0911.2483.
[7] T. Nikolaus and K. Waldorf, “Lifting problems and
this second step was performed in [6][theorem 99], transgression for non-abelian gerbes,” Adv. Math.
[40][theorem 5.1.29]. An analogous analysis for more 242, 50–79 (2013), arXiv: 1112.4702.
[8] J. Baez, A. Crans, D. Stevenson, and U. Schreiber,
general cases is beyond the scope of this paper, see in-
“From loop groups to 2-groups,” Homology, Homo-
stead [40][theorem 4.4.36]. See also [133] for a discussion topy Appl. 9, 101–135 (2007), math/0504123.
of bundles of 2-groups, [38, 39] for a discussion of more [9] T. Nikolaus, C. Sachse, and C. Wockel, “A smooth
general bundles of higher groups, and [7][section 3] for model for the String group,” Int. Math. Res. Not.
a discussion of the relationship between 2-groups and IMRN 16, 3678–3721 (2013), arXiv: 1104.4288.
1-gerbes. [10] D. Yetter, “TQFT’s from homotopy 2-types,” Jour-
Briefly, the result of the second step is that for G a nal of Knot Theory and its Ramifications 2, 113–123
(1993).
compact Lie group of dimension greater than zero, ex-
[11] J. Martins and T. Porter, “On Yetter’s invariants and
tensions of G by any flavor of Bk U(1) are classified by an extension of the Dijkgraaf-Witten invariant to cat-
H k+2 (G, Z), just as topological class of bundles. How- egorical groups,” Theor. Appl. Categor. 18, 118–150
ever, for G finite, there is additional28 information in (2007), math/0608484.
the second step, and classes of extensions with higher [12] H. Grosse and K.-G. Schlesinger, “Duals for non-
group structures are classified by group cohomology abelian lattice gauge theories by categorical meth-
H k+2 (G, U(1)) (with trivial action on the coefficients). ods,” Int. J. Theor. Phys. 40, 459–475 (2001).
[13] H. Pfeiffer, “Higher gauge theory and a non-
abelian generalization of 2-form electrodynamics,”
27 For G̃ and G̃ , this follows from [40][theorem 4.4.36]. For hep-th/0304074.
1,k 0,k
G̃c,k , this follows from unpublished work of D. Pavlov [27]. [14] J. Fröhlich, J. Fuchs, I. Runkel, and C. Schweigert,
The reader should also note that the geometric realization of “Defect lines, dualities, and generalised orbifolds,”
arXiv: 0909.5013.
Bk U(1)conn forgets the differential forms data, and so loses
[15] N. Carqueville and I. Runkel, “Orbifold completion
quite a bit of information. of defect bicategories,” arXiv: 1210.6363.
28 If G is finite, then for k ≥ 0, H k+2 (G, Z) vanishes, so the topo-
[16] N. Carqueville, A. R. Camacho, and I. Runkel, “Orb-
logical bundles are all trivial. However, one can still have a non- ifold equivalent potentials,” arXiv: 1311.3354.
trivial higher group extension, as this encodes additional struc- [17] I. Brunner, N. Carqueville, and D. Plencner, “Orb-
ture on top of the topological bundle. ifolds and topological defects,” arXiv: 1307.3141.

Fortschritte
der Physik
Progress
Original Paper
of Physics
[18] I. Brunner, N. Carqueville, and D. Plencner, “Discrete [41] D. Fiorenza, C. Rogers, and U. Schreiber, “A higher
torsion defects,” Comm. Math. Phys. 337, 429–453 Chern-Weil derivation of AKSZ sigma models,” Int.
(2015), arXiv: 1404.7497. J. Geom. Methods Mod. Phys. 10, 1250078 (2013),
[19] N. Carqueville and A. Q. Velez, “Calabi-Yau arXiv: 1108.4378.
completions and orbifold equivalences,” arXiv: [42] B. Rosenstein and A. Kovner, “Masslessness of pho-
1509.00880. ton and Goldstone theorem,” Int. J. Mod. Phys. A6,
[20] G. Moore, “Finite in all directions,” hep-th/9305139. 3559–3570 (1991).
[21] K. Waldorf, “A construction of string 2-group mod- [43] A. Kovner and B. Rosenstein, “New look at QED4 : the
els using a transgression-regression technique,” photon as a Goldstone boson and the topological in-
pp. 99–115 in Analysis, geometry, and quantum field terpretation of electric charge,” Phys. Rev. D49, 5571–
theory, Contemp. Math., vol. 584, (Amer. Math. Soc., 5582 (1994).
Providence, RI, 2012). [44] A. Kovner, “Confinement, magnetic ZN symmetry
[22] K. Waldorf, “Transgression to loop spaces and its in- and low energy effective theory of gluodynamics,”
verse, II: gerbes and fusion bundles with connectin,” pp. 1777–1825 in At the frontier of particle physics,
arXiv: 1004.0031. volume 3, ed. M. Shifman, (World Scientific, 2001),
[23] J. Baez and A. Lauda, “Higher-dimensional algebra hep-ph/0009138.
V: 2-groups,” math/0307200, Theor. Appl. Categor. 12, [45] C. K. Altes and A. Kovner, “Magnetic ZN symmetry
423–491 (2004). in hot QCD and the spatial Wilson loop,” Phys. Rev.
[24] J. Baez, A. Baratin, L. Freidel, and D. Wise, “Infinite- D62, 096008 (2000), hep-ph/0004052.
dimensional representations of 2-groups,” arXiv: [46] A. Kapustin and N. Saulina, “Topological boundary
0812.4969, Mem. Amer. Math. Soc. 219 (2012) no. conditions in abelian Chern-Simons theory,” arXiv:
1032. 1008.0654.
[25] H. Sinh, Gr-categories, (Université Paris VII doctoral [47] U. Schreiber, talk at Workshop on higher categories
thesis, 1975). and their applications (Fields Institute, Toronto, Jan-
[26] N. Ganter, private communication. uary 2007).
[27] U. Schreiber, private communication. [48] M. Berasaluce-González, G. Ramı́rez, and A. Uranga,
[28] J.-L. Brylinski, Loop spaces, characteristic classes, “Antisymmetric tensor Zp gauge symmetries in field
and geometric quantization (Birkhäuser, Boston, theory and string theory,” JHEP 1401, 059 (2014),
1993). arXiv: 1310.5582.
[29] E. Sharpe, “Discrete torsion,” Phys. Rev. D68, 126003 [49] M. Ando and E. Sharpe, “Two-dimensional topologi-
(2003), hep-th/0008154. cal field theories as taffy,” Adv. Theor. Math. Phys. 15,
[30] U. Bunke, T. Nikolaus, and M. Völkl, “Differential 179–244 (2011), arXiv: 1011.0100.
cohomology theories as sheaves of spectra,” arXiv: [50] T. Pantev and E. Sharpe, “GLSM’s for gerbes (and
1311.3188. other toric stacks),” Adv. Theor. Math. Phys. 10, 77–
[31] J. Milnor and J. Stasheff, Characteristic classes 121 (2006), hep-th/0502053.
(Princeton University Press, 1974). [51] T. Pantev and E. Sharpe, “Notes on gauging noneffec-
[32] D. Fiorenza, U. Schreiber, and J. Stasheff, “Cech tive group actions,” hep-th/0502027.
cocycles for differential characteristic classes – [52] T. Pantev and E. Sharpe, “String compactifications on
an infinity-Lie theoretic construction,” arXiv: Calabi-Yau stacks,” Nucl. Phys. B733, 233–296 (2006),
1011.4735. hep-th/0502044.
[33] P. Bressler, “The first Pontryagin class,” Compositio [53] O. Aharony, N. Seiberg, and Y. Tachikawa, “Reading
Math. 143, 1127–1163 (2007). between the lines of four-dimensional gauge theo-
[34] P. Bouwknegt, talk at String-Math 2011. ries,” arXiv: 1305.0318.
[35] M. Garcia-Fernandez, “Torsion-free generalized con- [54] D. Gaiotto, G. Moore, and A. Neitzke, “Framed BPS
nections and heterotic supergravity,” Comm. Math. states,” arXiv: 1006.0146.
Phys. 332, 89–115 (2014), arXiv: 1304.4294. [55] S. Hellerman and E. Sharpe, “Sums over topological
[36] D. Baraglia and P. Hekmati, “Transitive Courant al- sectors and quantization of Fayet-Iliopoulos param-
gebroids, string structures and T-duality,” arXiv: eters,” Adv. Theor. Math. Phys. 15, 1141–1199 (2011),
1308.5159. arXiv: 1012.5999.
[37] H. Sati, U. Schreiber and J. Stasheff, “Twisted dif- [56] S. Hellerman, A. Henriques, T. Pantev, E. Sharpe,
ferential String and Fivebrane structures,” Comm. and M. Ando, “Cluster decomposition, T-duality, and
Math. Phys. 315, 169–213 (2012), arXiv: 0910 gerby CFT’s,” Adv. Theor. Math. Phys. 11, 751–818
.4001. (2007), hep-th/0606034.
[38] T. Nikolaus, U. Schreiber and D. Stevenson, “Princi- [57] E. Sharpe, “Decomposition in diverse dimensions,”
pal ∞-bundles – general theory,” arXiv: 1207.0248. Phys. Rev. D90, 025030 (2014), arXiv: 1404.3986.
[39] T. Nikolaus, U. Schreiber, and D. Stevenson, “Princi- [58] L. Anderson, B. Jia, R. Manion, B. Ovrut, and
pal ∞-bundles – presentations,” arXiv: 1207.0249. E. Sharpe, “General aspects of heterotic string
[40] U. Schreiber, “Differential cohomology in a cohesive compactifications on stacks and gerbes,” arXiv:
∞-topos,” arXiv: 1310.7930. 1307.2269.
Fortschritte
der Physik
Fortschr. Phys. 63, No. 11–12 (2015)
Progress
Original Paper
of Physics
[59] E. Andreini, Y. Jiang, and H.-H. Tseng, “On Gromov- [80] N. Ganter and M. Kapranov, “Representation and
Witten theory of root gerbes,” arXiv: 0812.4477. character theory in 2-categories,” Adv. Math. 217,
[60] E. Andreini, Y. Jiang, and H.-H. Tseng, “Gromov- 2268–2300 (2008), math/0602510.
Witten theory of product stacks,” arXiv: 0905 [81] N. Ganter, “Categorical tori,” arXiv: 1406.7046.
.2258. [82] N. Epa, private communication.
[61] E. Andreini, Y. Jiang, and H.-H. Tseng, “Gromov Wit- [83] G. Moore and N. Seiberg, “Taming the conformal
ten theory of root gerbes I: structure of genus 0 mod- zoo,” Phys. Lett. B220, 422-430 (1989).
uli spaces,” arXiv: 0907.2087. [84] J. Lurie, “On the classification of topological
[62] E. Andreini, Y. Jiang, and H.-H. Tseng, “Gromov- field theories,” available at http://www.math
Witten theory of banded gerbes over schemes,” .harvard.edu/ lurie/papers/cobordism.pdf
arXiv: 1101.5996. [85] D. Fiorenza and A. Valentino, “Boundary con-
[63] A. Gholampour and H.-H. Tseng, “On Donaldson- ditions for topological quantum field theories,
Thomas invariants of threefold stacks and gerbes,” anomalies and projective modular functors,” arXiv:
Proc. Amer. Math. Soc. 141, 191–203 (2013). 1409.5723.
[64] H.-H. Tseng, “Notes on orbifold Gromov-Witten the- [86] D. Freed, “The cobordism hypothesis,” arXiv:
ory,” proceedings of the Fifth International Congress 1210.5100.
of Chinese Mathematicians, AMS/IP Stud. Adv. Math. [87] D. Freed and C. Teleman, “Relative quantum field
51, part 1, volume 2, pp. 327-340. theory,” Comm. Math. Phys. 326, 459–476 (2014),
[65] A. Caldararu, J. Distler, S. Hellerman, T. Pantev, and arXiv: 1212.1692.
E. Sharpe, “Non-birational twisted derived equiva- [88] D. Freed, “Anomalies and invertible field theories,”
lences in abelian GLSMs,” Comm. Math. Phys. 294, Proc. Symp. Pure Math. 88, 25–46 (2014), arXiv:
605–645 (2010), arXiv: 0709.3855. 1404.7224.
[66] K. Hori, “Duality in two-dimensional (2,2) supersym- [89] S. Weinberg, The quantum theory of fields, vol-
metric non-abelian gauge theories,” JHEP 1310, 121 ume II: modern applications (Cambridge University
(2013), arXiv: 1104.2853. Press, 1996).
[67] J. Halverson, V. Kumar, and D. Morrison, “New meth- [90] C. Itzykson and J.-B. Zuber, Quantum field theory
ods for characterizing phases of 2d supersymmet- (McGraw-Hill, 1980).
ric gauge theories,” JHEP 1309, 143 (2013), arXiv: [91] L. Faddeev, “Operator anomaly for the Gauss law,”
1305.3278. Phys. Lett. 145B, 81–84 (1984).
[68] E. Sharpe, “A few recent developments in 2d (2,2) [92] L. Faddeev and S. Shatashvili, “Algebraic and Hamil-
and (0,2) theories,” contribution to the proceedings tonian methods in the theory of nonabelian anoma-
of String-Math 2014, arXiv: 1501.01628. lies,” Theor. Math. Phys. 60, 770–778 (1985).
[69] R. Dijkgraaf and E. Witten, “Topological gauge the- [93] J. Mickelsson, “Chiral anomalies in even and odd di-
ories and group cohomology,” Comm. Math. Phys. mensions,” Comm. Math. Phys. 97, 361–370 (1985).
129, 393–429 (1990). [94] A. C Davis, J. A. Gracey, and A. J. Macfarlane,
[70] M. Douglas, “D-branes and discrete torsion,” “Anomalous current algebras in the Skyrme model
hep-th/9807235. or chiral G × G with Wess-Zumino term,” Phys. Lett.
[71] M. Douglas and B. Fiol, “D-branes and discrete tor- B194, 415–419 (1987).
sion 2,” JHEP 0509, 053 (2005), hep-th/9903031. [95] D. Freed, “Pions and generalized cohomology,”
[72] P. Aspinwall, “A note on the equivalence of Vafa’s and hep-th/0607134.
Douglas’s picture of discrete torsion,” JHEP 0012, 029 [96] E. Witten, “Global aspects of current algebra,” Nucl.
(2000), hep-th/0009045. Phys. B223, 422–432 (1983).
[73] T. Pantev, private communication. [97] L.-L. Chau, M.-L. Ge, and Y.-S. Wu, “Kac-Moody alge-
[74] E. Sharpe, “Analogues of discrete torsion for the M bra in the self-dual Yang-Mills equation,” Phys. Rev.
theory three-form,” Phys. Rev. D68, 126004 (2003), D25, 1086–1094 (1982).
hep-th/0008170. [98] L.-L. Chau and Y.-S. Wu, “More about hidden sym-
[75] S. Monnier, “Higher abelian Dijkgraaf-Witten the- metry algebra for the self-dual Yang-Mills equation,”
ory,” arXiv: 1502.04706. Phys. Rev. D26, 3581–3592 (1982).
[76] E. Witten, “Nonabelian bosonization in two dimen- [99] L.-L. Chau, M.-L. Ge, A. Sinha, and Y.-S. Wu, “Hidden
sions,” Comm. Math. Phys. 92, 455–472 (1984). symmetry algebra for the self-dual Yang-Mills equa-
[77] D. Fiorenza, H. Sati and U. Schreiber, “A higher tion,” Phys. Lett. B121, 391–396 (1983).
stacky perspective on Chern-Simons theory,” [100] L. Dolan, “A new symmetric group of real self-dual
pp. 153-211 in Mathematical Aspects of Quantum Yang-Mills theory,” Phys. Lett. B113, 387–390 (1982).
Field Theories, ed. D. Calaque, T. Strobl, (Springer, [101] L. Crane, “Action of the loop group on the self dual
2015), arXiv: 1301.2580. Yang-Mills equation,” Comm. Math. Phys. 110, 391–
[78] D. Gepner and E. Witten, “String theory on group 414 (1987).
manifolds,” Nucl. Phys. B278, 493-549 (1986). [102] T. He, P. Mitra, and A. Strominger, “2d Kac-
[79] U. Schreiber, “AQFT from n-functorial QFT,” Comm. Moody symmetry of 4d Yang-Mills theory,” arXiv:
Math. Phys. 291, 357–401 (2009), arXiv: 0806.1079. 1503.02663.

Fortschritte
der Physik
Progress
Original Paper
of Physics
[103] T. Adamo and E. Casali, “Perturbative gauge theory [119] S. J. Gates Jr. and W. Siegel, “Leftons, rightons, non-
at null infinity,” arXiv: 1504.02304. linear sigma models, and superstrings,” Phys. Lett.
[104] C. Vafa and E. Witten, “A strong coupling test B206, 631–638 (1988).
of S duality,” Nucl. Phys. B431, 3–77 (1994), [120] D. Depireux, S. J. Gates Jr., and Q-H. Park, “Lefton-
hep-th/9408074. righton formulation of massless Thirring models,”
[105] H. Nakajima, “Instantons on ALE spaces, quiver va- Phys. Lett B224, 364–372 (1989).
rieties, and Kac-Moody algebras,” Duke Math. J. 76, [121] S. Bellucci, D. Depireux, and S. J. Gates Jr., “(1, 0)
365–416 (1994). Thirring models and the coupling of spin-zero fields
[106] M. Henneaux and V. Lekeu, “Kac-Moody and to the heterotic string,” Phys. Lett B232, 67–74 (1989).
Borcherds symmetries of six-dimensional chiral su- [122] S. J. Gates Jr., S. Ketov, S. Kuzenko, and O. Soloviev,
pergravity,” arXiv: 1502.00518. “Lagrangian chiral coset construction of heterotic
[107] B. Julia, “Group disintegrations,” in Superspace string theories in (1, 0) superspace,” Nucl. Phys.
and supergravity (Nuffield Gravity Workshop, Cam- B362, 199–231 (1991).
bridge, England, June 22 - July 12, 1980), eds. S. W. [123] S. J. Gates Jr., “Strings, superstrings, and two-
Hawking and M. Rocek (Cambridge University Press, dimensional lagrangian field theory,” pp 140-184 in
Cambridge, 1981). Functional integration, geometry, and strings, pro-
[108] B. Julia, “Infinite Lie algebras in physics,” in Current ceedings of the XXV Winter School of Theoretical
Problems in Particle Physics, Unified Theories and Physics, Karpacz, Poland (Feb. 1989), ed. Z. Haba, J.
Beyond (John Hopkins University workshop, 1984). Sobczyk, Birkhauser, 1989.
[109] H. Nicolai, “A hyperbolic Lie algebra from supergrav- [124] L. Alvarez-Gaumé and P. Ginsparg, “The structure of
ity,” Phys. Lett. B276, 333–340 (1992). gauge and gravitational anomalies,” Ann. Phys. 161,
[110] B. Julia, “Dualities in the classical supergravity lim- 423–490 (1985).
its: dualizations, dualities, and a detour via (4k + 2)- [125] N. Ganter, “Hecke operators in equivariant elliptic
dimensions,” pp. 121-139 in Strings, branes, and du- cohomology and generalized Moonshine,” pp. 173-
alities (Cargese, 1997), hep-th/9805083. 209 in Groups and symmetries, CRM Proc. Lecture
[111] A. Vilenkin, E. Shellard, Cosmic Strings and Other Notes vol. 47, Amer. Math. Soc., Providence, Rhode
Topological Defects (Cambridge University Press, Island, 2009.
Cambridge, 1994). [126] T. Eguchi, H. Ooguri, and Y. Tachikawa, “Notes on the
[112] J. Preskill and L. Krauss, “Local discrete symmetry K3 surface and the Mathieu group M24 ,” Exper. Math.
and quantum mechanical hair,” Nucl. Phys. B341, 20, 91–96 (2011), arXiv: 1004.0956.
50–100 (1990). [127] T. Eguchi and K. Hikami, “Enriques moonshine,” J.
[113] A. Schwarz, “Field theories with no local conserva- Phys. A46, 312001 (2013), arXiv: 1301.5043.
tion of the electric charge,” Nucl. Phys. B208, 141– [128] T. Eguchi and K. Hikami, “N=2 moonshine,” Phys.
158 (1982). Lett. B717, 266–273 (2012), arXiv: 1209.0610.
[114] A. Schwarz and Y. Tyupkin, “Grand unification and [129] A. Taormina and K. Wendland, “The overarching fi-
mirror particles,” Nucl. Phys. B209, 427–432 (1982). nite symmetry group of Kummer surfaces in the
[115] M. Bucher, H.-K. Lo, and J. Preskill, “Topological ap- Mathieu group M24 ,” JHEP 1308, 125 (2013), arXiv:
proach to Alice electrodynamics,” Nucl. Phys. B386, 1107.3834.
3–26 (1992). [130] A. Taormina and K. Wendland, “The symmetries
[116] W.-K. Tung, Group Theory in Physics, (World Scien- of the tetrahedral Kummer surface in the Mathieu
tific, Philadelphia, PA, 1985). group M24 ,” arXiv: 1008.0954.
[117] J.-L. Brylinski and D. McLaughlin, “The geometry of [131] J. Conway, N. Elkies, and J. Martin, “The Mathieu
degree-four characteristic classes and of line bun- group M12 and its pseudogroup extension M13 ,” Ex-
dles on loop spaces I,” Duke Math. J. 75, 603–638 periment. Math. 15, 223–236 (2006).
(1994). [132] G. Moore and N. Seiberg, “Classical and quantum
[118] J. Distler and E. Sharpe, “Heterotic compactifica- conformal field theory,” Comm. Math. Phys. 123,
tions with principal bundles for general groups and 177–254 (1989).
general levels,” Adv. Theor. Math. Phys. 14, 335–398 [133] T. Bartels, “Higher gauge theory I: 2-bundles,”
(2010), hep-th/0701244. math/0410328.

Fortschritte Der Physik - 2015 - Sharpe - Notes On Generalized Global Symmetries in QFT

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fortschritte Der Physik - 2015 - Sharpe - Notes On Generalized Global Symmetries in QFT

Uploaded by

Copyright:

Available Formats

Fortschritte

Notes on generalized global symmetries in QFT

Received 4 September 2015, accepted 4 September 2015

There are more 2-groups and higher groups than

In this section, to make this paper self-contained, we ∼

Similarly, the identity axioms

2 Technically, we should describe this in terms of tensor prod-

space. (The resulting description will also implicitly en- of Q by

β(w, x, y, z) : α(w, x, y · z) ◦ α(w · x, y, z)

To make our notation uniform, we shall define B0 G to

and where D4 acts on X by first projecting to Z2 × Z2 , Since the Z2 acts trivially on X ,

boundaries is given in [69][equ’n (6.35)] as

α(g, h, k)α(h, k, g)α(k, g, h)

where subscripts indicate levels. 3.5 Current algebras

4 Cosmological defects and generalized

the Wilson line to be determined in part by the cos-

has stabilizer S(O(2) × O(1)) = O(2) ⊂ SO(3), so with that

The analogue for Wilson lines is similar:

The map π5 (SU(3)) → π4 (B3 U(1)) should be propor-

(In particular, since the geometric realizations of all three

You might also like